4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
25 /* Portions Copyright 2007 Jeremy Teo */
27 #include <sys/types.h>
28 #include <sys/param.h>
30 #include <sys/systm.h>
31 #include <sys/sysmacros.h>
32 #include <sys/resource.h>
34 #include <sys/vnode.h>
38 #include <sys/taskq.h>
40 #include <sys/atomic.h>
41 #include <sys/namei.h>
43 #include <sys/cmn_err.h>
44 #include <sys/errno.h>
45 #include <sys/unistd.h>
46 #include <sys/zfs_dir.h>
47 #include <sys/zfs_ioctl.h>
48 #include <sys/fs/zfs.h>
54 #include <sys/dirent.h>
55 #include <sys/policy.h>
56 #include <sys/sunddi.h>
57 #include <sys/filio.h>
59 #include <sys/zfs_ctldir.h>
60 #include <sys/zfs_fuid.h>
62 #include <sys/zfs_rlock.h>
63 #include <sys/extdirent.h>
64 #include <sys/kidmap.h>
67 #include <sys/sf_buf.h>
68 #include <sys/sched.h>
74 * Each vnode op performs some logical unit of work. To do this, the ZPL must
75 * properly lock its in-core state, create a DMU transaction, do the work,
76 * record this work in the intent log (ZIL), commit the DMU transaction,
77 * and wait for the intent log to commit if it is a synchronous operation.
78 * Moreover, the vnode ops must work in both normal and log replay context.
79 * The ordering of events is important to avoid deadlocks and references
80 * to freed memory. The example below illustrates the following Big Rules:
82 * (1) A check must be made in each zfs thread for a mounted file system.
83 * This is done avoiding races using ZFS_ENTER(zfsvfs).
84 * A ZFS_EXIT(zfsvfs) is needed before all returns. Any znodes
85 * must be checked with ZFS_VERIFY_ZP(zp). Both of these macros
86 * can return EIO from the calling function.
88 * (2) VN_RELE() should always be the last thing except for zil_commit()
89 * (if necessary) and ZFS_EXIT(). This is for 3 reasons:
90 * First, if it's the last reference, the vnode/znode
91 * can be freed, so the zp may point to freed memory. Second, the last
92 * reference will call zfs_zinactive(), which may induce a lot of work --
93 * pushing cached pages (which acquires range locks) and syncing out
94 * cached atime changes. Third, zfs_zinactive() may require a new tx,
95 * which could deadlock the system if you were already holding one.
96 * If you must call VN_RELE() within a tx then use VN_RELE_ASYNC().
98 * (3) All range locks must be grabbed before calling dmu_tx_assign(),
99 * as they can span dmu_tx_assign() calls.
101 * (4) Always pass TXG_NOWAIT as the second argument to dmu_tx_assign().
102 * This is critical because we don't want to block while holding locks.
103 * Note, in particular, that if a lock is sometimes acquired before
104 * the tx assigns, and sometimes after (e.g. z_lock), then failing to
105 * use a non-blocking assign can deadlock the system. The scenario:
107 * Thread A has grabbed a lock before calling dmu_tx_assign().
108 * Thread B is in an already-assigned tx, and blocks for this lock.
109 * Thread A calls dmu_tx_assign(TXG_WAIT) and blocks in txg_wait_open()
110 * forever, because the previous txg can't quiesce until B's tx commits.
112 * If dmu_tx_assign() returns ERESTART and zfsvfs->z_assign is TXG_NOWAIT,
113 * then drop all locks, call dmu_tx_wait(), and try again.
115 * (5) If the operation succeeded, generate the intent log entry for it
116 * before dropping locks. This ensures that the ordering of events
117 * in the intent log matches the order in which they actually occurred.
118 * During ZIL replay the zfs_log_* functions will update the sequence
119 * number to indicate the zil transaction has replayed.
121 * (6) At the end of each vnode op, the DMU tx must always commit,
122 * regardless of whether there were any errors.
124 * (7) After dropping all locks, invoke zil_commit(zilog, seq, foid)
125 * to ensure that synchronous semantics are provided when necessary.
127 * In general, this is how things should be ordered in each vnode op:
129 * ZFS_ENTER(zfsvfs); // exit if unmounted
131 * zfs_dirent_lock(&dl, ...) // lock directory entry (may VN_HOLD())
132 * rw_enter(...); // grab any other locks you need
133 * tx = dmu_tx_create(...); // get DMU tx
134 * dmu_tx_hold_*(); // hold each object you might modify
135 * error = dmu_tx_assign(tx, TXG_NOWAIT); // try to assign
137 * rw_exit(...); // drop locks
138 * zfs_dirent_unlock(dl); // unlock directory entry
139 * VN_RELE(...); // release held vnodes
140 * if (error == ERESTART) {
145 * dmu_tx_abort(tx); // abort DMU tx
146 * ZFS_EXIT(zfsvfs); // finished in zfs
147 * return (error); // really out of space
149 * error = do_real_work(); // do whatever this VOP does
151 * zfs_log_*(...); // on success, make ZIL entry
152 * dmu_tx_commit(tx); // commit DMU tx -- error or not
153 * rw_exit(...); // drop locks
154 * zfs_dirent_unlock(dl); // unlock directory entry
155 * VN_RELE(...); // release held vnodes
156 * zil_commit(zilog, seq, foid); // synchronous when necessary
157 * ZFS_EXIT(zfsvfs); // finished in zfs
158 * return (error); // done, report error
163 zfs_open(vnode_t **vpp, int flag, cred_t *cr, caller_context_t *ct)
165 znode_t *zp = VTOZ(*vpp);
166 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
171 if ((flag & FWRITE) && (zp->z_phys->zp_flags & ZFS_APPENDONLY) &&
172 ((flag & FAPPEND) == 0)) {
177 if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan &&
178 ZTOV(zp)->v_type == VREG &&
179 !(zp->z_phys->zp_flags & ZFS_AV_QUARANTINED) &&
180 zp->z_phys->zp_size > 0) {
181 if (fs_vscan(*vpp, cr, 0) != 0) {
187 /* Keep a count of the synchronous opens in the znode */
188 if (flag & (FSYNC | FDSYNC))
189 atomic_inc_32(&zp->z_sync_cnt);
197 zfs_close(vnode_t *vp, int flag, int count, offset_t offset, cred_t *cr,
198 caller_context_t *ct)
200 znode_t *zp = VTOZ(vp);
201 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
204 * Clean up any locks held by this process on the vp.
206 cleanlocks(vp, ddi_get_pid(), 0);
207 cleanshares(vp, ddi_get_pid());
212 /* Decrement the synchronous opens in the znode */
213 if ((flag & (FSYNC | FDSYNC)) && (count == 1))
214 atomic_dec_32(&zp->z_sync_cnt);
216 if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan &&
217 ZTOV(zp)->v_type == VREG &&
218 !(zp->z_phys->zp_flags & ZFS_AV_QUARANTINED) &&
219 zp->z_phys->zp_size > 0)
220 VERIFY(fs_vscan(vp, cr, 1) == 0);
227 * Lseek support for finding holes (cmd == _FIO_SEEK_HOLE) and
228 * data (cmd == _FIO_SEEK_DATA). "off" is an in/out parameter.
231 zfs_holey(vnode_t *vp, u_long cmd, offset_t *off)
233 znode_t *zp = VTOZ(vp);
234 uint64_t noff = (uint64_t)*off; /* new offset */
239 file_sz = zp->z_phys->zp_size;
240 if (noff >= file_sz) {
244 if (cmd == _FIO_SEEK_HOLE)
249 error = dmu_offset_next(zp->z_zfsvfs->z_os, zp->z_id, hole, &noff);
252 if ((error == ESRCH) || (noff > file_sz)) {
254 * Handle the virtual hole at the end of file.
271 zfs_ioctl(vnode_t *vp, u_long com, intptr_t data, int flag, cred_t *cred,
272 int *rvalp, caller_context_t *ct)
284 * The following two ioctls are used by bfu. Faking out,
285 * necessary to avoid bfu errors.
293 if (ddi_copyin((void *)data, &off, sizeof (off), flag))
297 zfsvfs = zp->z_zfsvfs;
301 /* offset parameter is in/out */
302 error = zfs_holey(vp, com, &off);
306 if (ddi_copyout(&off, (void *)data, sizeof (off), flag))
314 page_lookup(vnode_t *vp, int64_t start, int64_t off, int64_t nbytes)
320 VM_OBJECT_LOCK_ASSERT(obj, MA_OWNED);
323 if ((pp = vm_page_lookup(obj, OFF_TO_IDX(start))) != NULL &&
324 vm_page_is_valid(pp, (vm_offset_t)off, nbytes)) {
325 if ((pp->oflags & VPO_BUSY) != 0) {
327 * Reference the page before unlocking and
328 * sleeping so that the page daemon is less
329 * likely to reclaim it.
331 vm_page_lock_queues();
332 vm_page_flag_set(pp, PG_REFERENCED);
333 vm_page_sleep(pp, "zfsmwb");
337 vm_page_lock_queues();
339 vm_page_unlock_queues();
341 if (__predict_false(obj->cache != NULL)) {
342 vm_page_cache_free(obj, OFF_TO_IDX(start),
343 OFF_TO_IDX(start) + 1);
353 page_unlock(vm_page_t pp)
360 zfs_map_page(vm_page_t pp, struct sf_buf **sfp)
363 *sfp = sf_buf_alloc(pp, 0);
364 return ((caddr_t)sf_buf_kva(*sfp));
368 zfs_unmap_page(struct sf_buf *sf)
376 * When a file is memory mapped, we must keep the IO data synchronized
377 * between the DMU cache and the memory mapped pages. What this means:
379 * On Write: If we find a memory mapped page, we write to *both*
380 * the page and the dmu buffer.
384 update_pages(vnode_t *vp, int64_t start, int len, objset_t *os, uint64_t oid,
385 int segflg, dmu_tx_t *tx)
391 ASSERT(vp->v_mount != NULL);
395 off = start & PAGEOFFSET;
397 for (start &= PAGEMASK; len > 0; start += PAGESIZE) {
399 int nbytes = MIN(PAGESIZE - off, len);
401 if ((pp = page_lookup(vp, start, off, nbytes)) != NULL) {
404 VM_OBJECT_UNLOCK(obj);
405 va = zfs_map_page(pp, &sf);
406 if (segflg == UIO_NOCOPY) {
407 (void) dmu_write(os, oid, start+off, nbytes,
410 (void) dmu_read(os, oid, start+off, nbytes,
411 va+off, DMU_READ_PREFETCH);;
421 VM_OBJECT_UNLOCK(obj);
425 * When a file is memory mapped, we must keep the IO data synchronized
426 * between the DMU cache and the memory mapped pages. What this means:
428 * On Read: We "read" preferentially from memory mapped pages,
429 * else we default from the dmu buffer.
431 * NOTE: We will always "break up" the IO into PAGESIZE uiomoves when
432 * the file is memory mapped.
435 mappedread(vnode_t *vp, int nbytes, uio_t *uio)
437 znode_t *zp = VTOZ(vp);
438 objset_t *os = zp->z_zfsvfs->z_os;
449 ASSERT(vp->v_mount != NULL);
453 start = uio->uio_loffset;
454 off = start & PAGEOFFSET;
457 for (start &= PAGEMASK; len > 0; start += PAGESIZE) {
458 int bytes = MIN(PAGESIZE - off, len);
461 if ((m = vm_page_lookup(obj, OFF_TO_IDX(start))) != NULL &&
462 vm_page_is_valid(m, off, bytes)) {
463 if ((m->oflags & VPO_BUSY) != 0) {
465 * Reference the page before unlocking and
466 * sleeping so that the page daemon is less
467 * likely to reclaim it.
469 vm_page_lock_queues();
470 vm_page_flag_set(m, PG_REFERENCED);
471 vm_page_sleep(m, "zfsmrb");
476 VM_OBJECT_UNLOCK(obj);
478 error = dmu_read_uio(os, zp->z_id, uio,
483 uiomove_fromphys(&m, off, bytes, uio);
486 } else if (m != NULL && uio->uio_segflg == UIO_NOCOPY) {
488 * The code below is here to make sendfile(2) work
489 * correctly with ZFS. As pointed out by ups@
490 * sendfile(2) should be changed to use VOP_GETPAGES(),
491 * but it pessimize performance of sendfile/UFS, that's
492 * why I handle this special case in ZFS code.
495 ("unexpected offset in mappedread for sendfile"));
496 if ((m->oflags & VPO_BUSY) != 0) {
498 * Reference the page before unlocking and
499 * sleeping so that the page daemon is less
500 * likely to reclaim it.
502 vm_page_lock_queues();
503 vm_page_flag_set(m, PG_REFERENCED);
504 vm_page_sleep(m, "zfsmrb");
508 VM_OBJECT_UNLOCK(obj);
510 error = dmu_read_uio(os, zp->z_id, uio,
515 va = zfs_map_page(m, &sf);
516 error = dmu_read(os, zp->z_id, start, bytes, va,
518 if (bytes != PAGE_SIZE)
519 bzero(va + bytes, PAGE_SIZE - bytes);
524 m->valid = VM_PAGE_BITS_ALL;
527 uio->uio_resid -= bytes;
528 uio->uio_offset += bytes;
538 VM_OBJECT_UNLOCK(obj);
539 if (error == 0 && dirbytes > 0)
540 error = dmu_read_uio(os, zp->z_id, uio, dirbytes);
544 offset_t zfs_read_chunk_size = 1024 * 1024; /* Tunable */
547 * Read bytes from specified file into supplied buffer.
549 * IN: vp - vnode of file to be read from.
550 * uio - structure supplying read location, range info,
552 * ioflag - SYNC flags; used to provide FRSYNC semantics.
553 * cr - credentials of caller.
554 * ct - caller context
556 * OUT: uio - updated offset and range, buffer filled.
558 * RETURN: 0 if success
559 * error code if failure
562 * vp - atime updated if byte count > 0
566 zfs_read(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct)
568 znode_t *zp = VTOZ(vp);
569 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
579 if (zp->z_phys->zp_flags & ZFS_AV_QUARANTINED) {
585 * Validate file offset
587 if (uio->uio_loffset < (offset_t)0) {
593 * Fasttrack empty reads
595 if (uio->uio_resid == 0) {
601 * Check for mandatory locks
603 if (MANDMODE((mode_t)zp->z_phys->zp_mode)) {
604 if (error = chklock(vp, FREAD,
605 uio->uio_loffset, uio->uio_resid, uio->uio_fmode, ct)) {
612 * If we're in FRSYNC mode, sync out this znode before reading it.
615 zil_commit(zfsvfs->z_log, zp->z_last_itx, zp->z_id);
618 * Lock the range against changes.
620 rl = zfs_range_lock(zp, uio->uio_loffset, uio->uio_resid, RL_READER);
623 * If we are reading past end-of-file we can skip
624 * to the end; but we might still need to set atime.
626 if (uio->uio_loffset >= zp->z_phys->zp_size) {
631 ASSERT(uio->uio_loffset < zp->z_phys->zp_size);
632 n = MIN(uio->uio_resid, zp->z_phys->zp_size - uio->uio_loffset);
635 nbytes = MIN(n, zfs_read_chunk_size -
636 P2PHASE(uio->uio_loffset, zfs_read_chunk_size));
638 if (vn_has_cached_data(vp))
639 error = mappedread(vp, nbytes, uio);
641 error = dmu_read_uio(os, zp->z_id, uio, nbytes);
643 /* convert checksum errors into IO errors */
653 zfs_range_unlock(rl);
655 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
661 * Fault in the pages of the first n bytes specified by the uio structure.
662 * 1 byte in each page is touched and the uio struct is unmodified.
663 * Any error will exit this routine as this is only a best
664 * attempt to get the pages resident. This is a copy of ufs_trans_touch().
667 zfs_prefault_write(ssize_t n, struct uio *uio)
673 if (uio->uio_segflg != UIO_USERSPACE)
679 cnt = MIN(iov->iov_len, n);
681 /* empty iov entry */
687 * touch each page in this segment.
693 incr = MIN(cnt, PAGESIZE);
698 * touch the last byte in case it straddles a page.
708 * Write the bytes to a file.
710 * IN: vp - vnode of file to be written to.
711 * uio - structure supplying write location, range info,
713 * ioflag - IO_APPEND flag set if in append mode.
714 * cr - credentials of caller.
715 * ct - caller context (NFS/CIFS fem monitor only)
717 * OUT: uio - updated offset and range.
719 * RETURN: 0 if success
720 * error code if failure
723 * vp - ctime|mtime updated if byte count > 0
727 zfs_write(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct)
729 znode_t *zp = VTOZ(vp);
730 rlim64_t limit = MAXOFFSET_T;
731 ssize_t start_resid = uio->uio_resid;
735 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
740 int max_blksz = zfsvfs->z_max_blksz;
746 * Fasttrack empty write
752 if (limit == RLIM64_INFINITY || limit > MAXOFFSET_T)
759 * If immutable or not appending then return EPERM
761 pflags = zp->z_phys->zp_flags;
762 if ((pflags & (ZFS_IMMUTABLE | ZFS_READONLY)) ||
763 ((pflags & ZFS_APPENDONLY) && !(ioflag & FAPPEND) &&
764 (uio->uio_loffset < zp->z_phys->zp_size))) {
769 zilog = zfsvfs->z_log;
772 * Pre-fault the pages to ensure slow (eg NFS) pages
775 zfs_prefault_write(n, uio);
778 * If in append mode, set the io offset pointer to eof.
780 if (ioflag & IO_APPEND) {
782 * Range lock for a file append:
783 * The value for the start of range will be determined by
784 * zfs_range_lock() (to guarantee append semantics).
785 * If this write will cause the block size to increase,
786 * zfs_range_lock() will lock the entire file, so we must
787 * later reduce the range after we grow the block size.
789 rl = zfs_range_lock(zp, 0, n, RL_APPEND);
790 if (rl->r_len == UINT64_MAX) {
791 /* overlocked, zp_size can't change */
792 woff = uio->uio_loffset = zp->z_phys->zp_size;
794 woff = uio->uio_loffset = rl->r_off;
797 woff = uio->uio_loffset;
799 * Validate file offset
807 * If we need to grow the block size then zfs_range_lock()
808 * will lock a wider range than we request here.
809 * Later after growing the block size we reduce the range.
811 rl = zfs_range_lock(zp, woff, n, RL_WRITER);
815 zfs_range_unlock(rl);
820 if ((woff + n) > limit || woff > (limit - n))
824 * Check for mandatory locks
826 if (MANDMODE((mode_t)zp->z_phys->zp_mode) &&
827 (error = chklock(vp, FWRITE, woff, n, uio->uio_fmode, ct)) != 0) {
828 zfs_range_unlock(rl);
832 end_size = MAX(zp->z_phys->zp_size, woff + n);
835 * Write the file in reasonable size chunks. Each chunk is written
836 * in a separate transaction; this keeps the intent log records small
837 * and allows us to do more fine-grained space accounting.
841 woff = uio->uio_loffset;
844 if (zfs_usergroup_overquota(zfsvfs,
845 B_FALSE, zp->z_phys->zp_uid) ||
846 zfs_usergroup_overquota(zfsvfs,
847 B_TRUE, zp->z_phys->zp_gid)) {
849 dmu_return_arcbuf(abuf);
855 * If dmu_assign_arcbuf() is expected to execute with minimum
856 * overhead loan an arc buffer and copy user data to it before
857 * we enter a txg. This avoids holding a txg forever while we
858 * pagefault on a hanging NFS server mapping.
860 if (abuf == NULL && n >= max_blksz &&
861 woff >= zp->z_phys->zp_size &&
862 P2PHASE(woff, max_blksz) == 0 &&
863 zp->z_blksz == max_blksz) {
866 abuf = dmu_request_arcbuf(zp->z_dbuf, max_blksz);
867 ASSERT(abuf != NULL);
868 ASSERT(arc_buf_size(abuf) == max_blksz);
869 if (error = uiocopy(abuf->b_data, max_blksz,
870 UIO_WRITE, uio, &cbytes)) {
871 dmu_return_arcbuf(abuf);
874 ASSERT(cbytes == max_blksz);
878 * Start a transaction.
880 tx = dmu_tx_create(zfsvfs->z_os);
881 dmu_tx_hold_bonus(tx, zp->z_id);
882 dmu_tx_hold_write(tx, zp->z_id, woff, MIN(n, max_blksz));
883 error = dmu_tx_assign(tx, TXG_NOWAIT);
885 if (error == ERESTART) {
892 dmu_return_arcbuf(abuf);
897 * If zfs_range_lock() over-locked we grow the blocksize
898 * and then reduce the lock range. This will only happen
899 * on the first iteration since zfs_range_reduce() will
900 * shrink down r_len to the appropriate size.
902 if (rl->r_len == UINT64_MAX) {
905 if (zp->z_blksz > max_blksz) {
906 ASSERT(!ISP2(zp->z_blksz));
907 new_blksz = MIN(end_size, SPA_MAXBLOCKSIZE);
909 new_blksz = MIN(end_size, max_blksz);
911 zfs_grow_blocksize(zp, new_blksz, tx);
912 zfs_range_reduce(rl, woff, n);
916 * XXX - should we really limit each write to z_max_blksz?
917 * Perhaps we should use SPA_MAXBLOCKSIZE chunks?
919 nbytes = MIN(n, max_blksz - P2PHASE(woff, max_blksz));
921 if (woff + nbytes > zp->z_phys->zp_size)
922 vnode_pager_setsize(vp, woff + nbytes);
925 tx_bytes = uio->uio_resid;
926 error = dmu_write_uio(zfsvfs->z_os, zp->z_id, uio,
928 tx_bytes -= uio->uio_resid;
931 ASSERT(tx_bytes == max_blksz);
932 dmu_assign_arcbuf(zp->z_dbuf, woff, abuf, tx);
933 ASSERT(tx_bytes <= uio->uio_resid);
934 uioskip(uio, tx_bytes);
937 if (tx_bytes && vn_has_cached_data(vp)) {
938 update_pages(vp, woff, tx_bytes, zfsvfs->z_os,
939 zp->z_id, uio->uio_segflg, tx);
943 * If we made no progress, we're done. If we made even
944 * partial progress, update the znode and ZIL accordingly.
953 * Clear Set-UID/Set-GID bits on successful write if not
954 * privileged and at least one of the excute bits is set.
956 * It would be nice to to this after all writes have
957 * been done, but that would still expose the ISUID/ISGID
958 * to another app after the partial write is committed.
960 * Note: we don't call zfs_fuid_map_id() here because
961 * user 0 is not an ephemeral uid.
963 mutex_enter(&zp->z_acl_lock);
964 if ((zp->z_phys->zp_mode & (S_IXUSR | (S_IXUSR >> 3) |
965 (S_IXUSR >> 6))) != 0 &&
966 (zp->z_phys->zp_mode & (S_ISUID | S_ISGID)) != 0 &&
967 secpolicy_vnode_setid_retain(vp, cr,
968 (zp->z_phys->zp_mode & S_ISUID) != 0 &&
969 zp->z_phys->zp_uid == 0) != 0) {
970 zp->z_phys->zp_mode &= ~(S_ISUID | S_ISGID);
972 mutex_exit(&zp->z_acl_lock);
975 * Update time stamp. NOTE: This marks the bonus buffer as
976 * dirty, so we don't have to do it again for zp_size.
978 zfs_time_stamper(zp, CONTENT_MODIFIED, tx);
981 * Update the file size (zp_size) if it has changed;
982 * account for possible concurrent updates.
984 while ((end_size = zp->z_phys->zp_size) < uio->uio_loffset)
985 (void) atomic_cas_64(&zp->z_phys->zp_size, end_size,
987 zfs_log_write(zilog, tx, TX_WRITE, zp, woff, tx_bytes, ioflag);
992 ASSERT(tx_bytes == nbytes);
996 zfs_range_unlock(rl);
999 * If we're in replay mode, or we made no progress, return error.
1000 * Otherwise, it's at least a partial write, so it's successful.
1002 if (zfsvfs->z_replay || uio->uio_resid == start_resid) {
1007 if (ioflag & (FSYNC | FDSYNC))
1008 zil_commit(zilog, zp->z_last_itx, zp->z_id);
1015 zfs_get_done(dmu_buf_t *db, void *vzgd)
1017 zgd_t *zgd = (zgd_t *)vzgd;
1018 rl_t *rl = zgd->zgd_rl;
1019 vnode_t *vp = ZTOV(rl->r_zp);
1020 objset_t *os = rl->r_zp->z_zfsvfs->z_os;
1023 vfslocked = VFS_LOCK_GIANT(vp->v_vfsp);
1024 dmu_buf_rele(db, vzgd);
1025 zfs_range_unlock(rl);
1027 * Release the vnode asynchronously as we currently have the
1028 * txg stopped from syncing.
1030 VN_RELE_ASYNC(vp, dsl_pool_vnrele_taskq(dmu_objset_pool(os)));
1031 zil_add_block(zgd->zgd_zilog, zgd->zgd_bp);
1032 kmem_free(zgd, sizeof (zgd_t));
1033 VFS_UNLOCK_GIANT(vfslocked);
1037 * Get data to generate a TX_WRITE intent log record.
1040 zfs_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio)
1042 zfsvfs_t *zfsvfs = arg;
1043 objset_t *os = zfsvfs->z_os;
1045 uint64_t off = lr->lr_offset;
1049 int dlen = lr->lr_length; /* length of user data */
1056 * Nothing to do if the file has been removed
1058 if (zfs_zget(zfsvfs, lr->lr_foid, &zp) != 0)
1060 if (zp->z_unlinked) {
1062 * Release the vnode asynchronously as we currently have the
1063 * txg stopped from syncing.
1065 VN_RELE_ASYNC(ZTOV(zp),
1066 dsl_pool_vnrele_taskq(dmu_objset_pool(os)));
1071 * Write records come in two flavors: immediate and indirect.
1072 * For small writes it's cheaper to store the data with the
1073 * log record (immediate); for large writes it's cheaper to
1074 * sync the data and get a pointer to it (indirect) so that
1075 * we don't have to write the data twice.
1077 if (buf != NULL) { /* immediate write */
1078 rl = zfs_range_lock(zp, off, dlen, RL_READER);
1079 /* test for truncation needs to be done while range locked */
1080 if (off >= zp->z_phys->zp_size) {
1084 VERIFY(0 == dmu_read(os, lr->lr_foid, off, dlen, buf,
1085 DMU_READ_NO_PREFETCH));
1086 } else { /* indirect write */
1087 uint64_t boff; /* block starting offset */
1090 * Have to lock the whole block to ensure when it's
1091 * written out and it's checksum is being calculated
1092 * that no one can change the data. We need to re-check
1093 * blocksize after we get the lock in case it's changed!
1096 if (ISP2(zp->z_blksz)) {
1097 boff = P2ALIGN_TYPED(off, zp->z_blksz,
1103 rl = zfs_range_lock(zp, boff, dlen, RL_READER);
1104 if (zp->z_blksz == dlen)
1106 zfs_range_unlock(rl);
1108 /* test for truncation needs to be done while range locked */
1109 if (off >= zp->z_phys->zp_size) {
1113 zgd = (zgd_t *)kmem_alloc(sizeof (zgd_t), KM_SLEEP);
1115 zgd->zgd_zilog = zfsvfs->z_log;
1116 zgd->zgd_bp = &lr->lr_blkptr;
1117 VERIFY(0 == dmu_buf_hold(os, lr->lr_foid, boff, zgd, &db));
1118 ASSERT(boff == db->db_offset);
1119 lr->lr_blkoff = off - boff;
1120 error = dmu_sync(zio, db, &lr->lr_blkptr,
1121 lr->lr_common.lrc_txg, zfs_get_done, zgd);
1122 ASSERT((error && error != EINPROGRESS) ||
1123 lr->lr_length <= zp->z_blksz);
1126 * dmu_sync() can compress a block of zeros to a null
1127 * blkptr but the block size still needs to be passed
1128 * through to replay.
1130 BP_SET_LSIZE(&lr->lr_blkptr, db->db_size);
1131 zil_add_block(zfsvfs->z_log, &lr->lr_blkptr);
1135 * If we get EINPROGRESS, then we need to wait for a
1136 * write IO initiated by dmu_sync() to complete before
1137 * we can release this dbuf. We will finish everything
1138 * up in the zfs_get_done() callback.
1140 if (error == EINPROGRESS) {
1142 } else if (error == EALREADY) {
1143 lr->lr_common.lrc_txtype = TX_WRITE2;
1146 dmu_buf_rele(db, zgd);
1147 kmem_free(zgd, sizeof (zgd_t));
1150 zfs_range_unlock(rl);
1152 * Release the vnode asynchronously as we currently have the
1153 * txg stopped from syncing.
1155 VN_RELE_ASYNC(ZTOV(zp), dsl_pool_vnrele_taskq(dmu_objset_pool(os)));
1161 zfs_access(vnode_t *vp, int mode, int flag, cred_t *cr,
1162 caller_context_t *ct)
1164 znode_t *zp = VTOZ(vp);
1165 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1171 if (flag & V_ACE_MASK)
1172 error = zfs_zaccess(zp, mode, flag, B_FALSE, cr);
1174 error = zfs_zaccess_rwx(zp, mode, flag, cr);
1181 * If vnode is for a device return a specfs vnode instead.
1184 specvp_check(vnode_t **vpp, cred_t *cr)
1188 if (IS_DEVVP(*vpp)) {
1191 svp = specvp(*vpp, (*vpp)->v_rdev, (*vpp)->v_type, cr);
1202 * Lookup an entry in a directory, or an extended attribute directory.
1203 * If it exists, return a held vnode reference for it.
1205 * IN: dvp - vnode of directory to search.
1206 * nm - name of entry to lookup.
1207 * pnp - full pathname to lookup [UNUSED].
1208 * flags - LOOKUP_XATTR set if looking for an attribute.
1209 * rdir - root directory vnode [UNUSED].
1210 * cr - credentials of caller.
1211 * ct - caller context
1212 * direntflags - directory lookup flags
1213 * realpnp - returned pathname.
1215 * OUT: vpp - vnode of located entry, NULL if not found.
1217 * RETURN: 0 if success
1218 * error code if failure
1225 zfs_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, struct componentname *cnp,
1226 int nameiop, cred_t *cr, kthread_t *td, int flags)
1228 znode_t *zdp = VTOZ(dvp);
1229 zfsvfs_t *zfsvfs = zdp->z_zfsvfs;
1231 int *direntflags = NULL;
1232 void *realpnp = NULL;
1235 if (!(flags & (LOOKUP_XATTR | FIGNORECASE))) {
1237 if (dvp->v_type != VDIR) {
1239 } else if (zdp->z_dbuf == NULL) {
1243 if (nm[0] == 0 || (nm[0] == '.' && nm[1] == '\0')) {
1244 error = zfs_fastaccesschk_execute(zdp, cr);
1252 vnode_t *tvp = dnlc_lookup(dvp, nm);
1255 error = zfs_fastaccesschk_execute(zdp, cr);
1260 if (tvp == DNLC_NO_VNODE) {
1265 return (specvp_check(vpp, cr));
1271 DTRACE_PROBE2(zfs__fastpath__lookup__miss, vnode_t *, dvp, char *, nm);
1278 if (flags & LOOKUP_XATTR) {
1281 * If the xattr property is off, refuse the lookup request.
1283 if (!(zfsvfs->z_vfs->vfs_flag & VFS_XATTR)) {
1290 * We don't allow recursive attributes..
1291 * Maybe someday we will.
1293 if (zdp->z_phys->zp_flags & ZFS_XATTR) {
1298 if (error = zfs_get_xattrdir(VTOZ(dvp), vpp, cr, flags)) {
1304 * Do we have permission to get into attribute directory?
1307 if (error = zfs_zaccess(VTOZ(*vpp), ACE_EXECUTE, 0,
1317 if (dvp->v_type != VDIR) {
1323 * Check accessibility of directory.
1326 if (error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr)) {
1331 if (zfsvfs->z_utf8 && u8_validate(nm, strlen(nm),
1332 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
1337 error = zfs_dirlook(zdp, nm, vpp, flags, direntflags, realpnp);
1339 error = specvp_check(vpp, cr);
1341 /* Translate errors and add SAVENAME when needed. */
1342 if (cnp->cn_flags & ISLASTCN) {
1346 if (error == ENOENT) {
1347 error = EJUSTRETURN;
1348 cnp->cn_flags |= SAVENAME;
1354 cnp->cn_flags |= SAVENAME;
1358 if (error == 0 && (nm[0] != '.' || nm[1] != '\0')) {
1361 if (cnp->cn_flags & ISDOTDOT) {
1362 ltype = VOP_ISLOCKED(dvp);
1366 error = vn_lock(*vpp, cnp->cn_lkflags);
1367 if (cnp->cn_flags & ISDOTDOT)
1368 vn_lock(dvp, ltype | LK_RETRY);
1378 #ifdef FREEBSD_NAMECACHE
1380 * Insert name into cache (as non-existent) if appropriate.
1382 if (error == ENOENT && (cnp->cn_flags & MAKEENTRY) && nameiop != CREATE)
1383 cache_enter(dvp, *vpp, cnp);
1385 * Insert name into cache if appropriate.
1387 if (error == 0 && (cnp->cn_flags & MAKEENTRY)) {
1388 if (!(cnp->cn_flags & ISLASTCN) ||
1389 (nameiop != DELETE && nameiop != RENAME)) {
1390 cache_enter(dvp, *vpp, cnp);
1399 * Attempt to create a new entry in a directory. If the entry
1400 * already exists, truncate the file if permissible, else return
1401 * an error. Return the vp of the created or trunc'd file.
1403 * IN: dvp - vnode of directory to put new file entry in.
1404 * name - name of new file entry.
1405 * vap - attributes of new file.
1406 * excl - flag indicating exclusive or non-exclusive mode.
1407 * mode - mode to open file with.
1408 * cr - credentials of caller.
1409 * flag - large file flag [UNUSED].
1410 * ct - caller context
1411 * vsecp - ACL to be set
1413 * OUT: vpp - vnode of created or trunc'd entry.
1415 * RETURN: 0 if success
1416 * error code if failure
1419 * dvp - ctime|mtime updated if new entry created
1420 * vp - ctime|mtime always, atime if new
1425 zfs_create(vnode_t *dvp, char *name, vattr_t *vap, int excl, int mode,
1426 vnode_t **vpp, cred_t *cr, kthread_t *td)
1428 znode_t *zp, *dzp = VTOZ(dvp);
1429 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1437 gid_t gid = crgetgid(cr);
1438 zfs_acl_ids_t acl_ids;
1439 boolean_t fuid_dirtied;
1444 * If we have an ephemeral id, ACL, or XVATTR then
1445 * make sure file system is at proper version
1448 ksid = crgetsid(cr, KSID_OWNER);
1450 uid = ksid_getid(ksid);
1453 if (zfsvfs->z_use_fuids == B_FALSE &&
1454 (vsecp || (vap->va_mask & AT_XVATTR) ||
1455 IS_EPHEMERAL(crgetuid(cr)) || IS_EPHEMERAL(crgetgid(cr))))
1461 zilog = zfsvfs->z_log;
1463 if (zfsvfs->z_utf8 && u8_validate(name, strlen(name),
1464 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
1469 if (vap->va_mask & AT_XVATTR) {
1470 if ((error = secpolicy_xvattr(dvp, (xvattr_t *)vap,
1471 crgetuid(cr), cr, vap->va_type)) != 0) {
1479 if ((vap->va_mode & S_ISVTX) && secpolicy_vnode_stky_modify(cr))
1480 vap->va_mode &= ~S_ISVTX;
1482 if (*name == '\0') {
1484 * Null component name refers to the directory itself.
1491 /* possible VN_HOLD(zp) */
1494 if (flag & FIGNORECASE)
1497 error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
1500 if (strcmp(name, "..") == 0)
1510 * Create a new file object and update the directory
1513 if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) {
1518 * We only support the creation of regular files in
1519 * extended attribute directories.
1521 if ((dzp->z_phys->zp_flags & ZFS_XATTR) &&
1522 (vap->va_type != VREG)) {
1528 if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, vsecp,
1531 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
1532 zfs_acl_ids_free(&acl_ids);
1537 tx = dmu_tx_create(os);
1538 dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
1539 fuid_dirtied = zfsvfs->z_fuid_dirty;
1541 zfs_fuid_txhold(zfsvfs, tx);
1542 dmu_tx_hold_bonus(tx, dzp->z_id);
1543 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
1544 if (acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) {
1545 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
1546 0, SPA_MAXBLOCKSIZE);
1548 error = dmu_tx_assign(tx, TXG_NOWAIT);
1550 zfs_acl_ids_free(&acl_ids);
1551 zfs_dirent_unlock(dl);
1552 if (error == ERESTART) {
1561 zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, &acl_ids);
1564 zfs_fuid_sync(zfsvfs, tx);
1566 (void) zfs_link_create(dl, zp, tx, ZNEW);
1568 txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap);
1569 if (flag & FIGNORECASE)
1571 zfs_log_create(zilog, tx, txtype, dzp, zp, name,
1572 vsecp, acl_ids.z_fuidp, vap);
1573 zfs_acl_ids_free(&acl_ids);
1576 int aflags = (flag & FAPPEND) ? V_APPEND : 0;
1579 * A directory entry already exists for this name.
1582 * Can't truncate an existing file if in exclusive mode.
1589 * Can't open a directory for writing.
1591 if ((ZTOV(zp)->v_type == VDIR) && (mode & S_IWRITE)) {
1596 * Verify requested access to file.
1598 if (mode && (error = zfs_zaccess_rwx(zp, mode, aflags, cr))) {
1602 mutex_enter(&dzp->z_lock);
1604 mutex_exit(&dzp->z_lock);
1607 * Truncate regular files if requested.
1609 if ((ZTOV(zp)->v_type == VREG) &&
1610 (vap->va_mask & AT_SIZE) && (vap->va_size == 0)) {
1611 /* we can't hold any locks when calling zfs_freesp() */
1612 zfs_dirent_unlock(dl);
1614 error = zfs_freesp(zp, 0, 0, mode, TRUE);
1616 vnevent_create(ZTOV(zp), ct);
1622 zfs_dirent_unlock(dl);
1629 error = specvp_check(vpp, cr);
1637 * Remove an entry from a directory.
1639 * IN: dvp - vnode of directory to remove entry from.
1640 * name - name of entry to remove.
1641 * cr - credentials of caller.
1642 * ct - caller context
1643 * flags - case flags
1645 * RETURN: 0 if success
1646 * error code if failure
1650 * vp - ctime (if nlink > 0)
1654 zfs_remove(vnode_t *dvp, char *name, cred_t *cr, caller_context_t *ct,
1657 znode_t *zp, *dzp = VTOZ(dvp);
1658 znode_t *xzp = NULL;
1660 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1662 uint64_t acl_obj, xattr_obj;
1665 boolean_t may_delete_now, delete_now = FALSE;
1666 boolean_t unlinked, toobig = FALSE;
1668 pathname_t *realnmp = NULL;
1675 zilog = zfsvfs->z_log;
1677 if (flags & FIGNORECASE) {
1685 * Attempt to lock directory; fail if entry doesn't exist.
1687 if (error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
1697 if (error = zfs_zaccess_delete(dzp, zp, cr)) {
1702 * Need to use rmdir for removing directories.
1704 if (vp->v_type == VDIR) {
1709 vnevent_remove(vp, dvp, name, ct);
1712 dnlc_remove(dvp, realnmp->pn_buf);
1714 dnlc_remove(dvp, name);
1716 may_delete_now = FALSE;
1719 * We may delete the znode now, or we may put it in the unlinked set;
1720 * it depends on whether we're the last link, and on whether there are
1721 * other holds on the vnode. So we dmu_tx_hold() the right things to
1722 * allow for either case.
1724 tx = dmu_tx_create(zfsvfs->z_os);
1725 dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name);
1726 dmu_tx_hold_bonus(tx, zp->z_id);
1727 if (may_delete_now) {
1729 zp->z_phys->zp_size > zp->z_blksz * DMU_MAX_DELETEBLKCNT;
1730 /* if the file is too big, only hold_free a token amount */
1731 dmu_tx_hold_free(tx, zp->z_id, 0,
1732 (toobig ? DMU_MAX_ACCESS : DMU_OBJECT_END));
1735 /* are there any extended attributes? */
1736 if ((xattr_obj = zp->z_phys->zp_xattr) != 0) {
1737 /* XXX - do we need this if we are deleting? */
1738 dmu_tx_hold_bonus(tx, xattr_obj);
1741 /* are there any additional acls */
1742 if ((acl_obj = zp->z_phys->zp_acl.z_acl_extern_obj) != 0 &&
1744 dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END);
1746 /* charge as an update -- would be nice not to charge at all */
1747 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
1749 error = dmu_tx_assign(tx, TXG_NOWAIT);
1751 zfs_dirent_unlock(dl);
1753 if (error == ERESTART) {
1766 * Remove the directory entry.
1768 error = zfs_link_destroy(dl, zp, tx, zflg, &unlinked);
1775 if (0 && unlinked) {
1777 delete_now = may_delete_now && !toobig &&
1778 vp->v_count == 1 && !vn_has_cached_data(vp) &&
1779 zp->z_phys->zp_xattr == xattr_obj &&
1780 zp->z_phys->zp_acl.z_acl_extern_obj == acl_obj;
1785 if (zp->z_phys->zp_xattr) {
1786 error = zfs_zget(zfsvfs, zp->z_phys->zp_xattr, &xzp);
1787 ASSERT3U(error, ==, 0);
1788 ASSERT3U(xzp->z_phys->zp_links, ==, 2);
1789 dmu_buf_will_dirty(xzp->z_dbuf, tx);
1790 mutex_enter(&xzp->z_lock);
1791 xzp->z_unlinked = 1;
1792 xzp->z_phys->zp_links = 0;
1793 mutex_exit(&xzp->z_lock);
1794 zfs_unlinked_add(xzp, tx);
1795 zp->z_phys->zp_xattr = 0; /* probably unnecessary */
1797 mutex_enter(&zp->z_lock);
1800 ASSERT3U(vp->v_count, ==, 0);
1802 mutex_exit(&zp->z_lock);
1803 zfs_znode_delete(zp, tx);
1804 } else if (unlinked) {
1805 zfs_unlinked_add(zp, tx);
1809 if (flags & FIGNORECASE)
1811 zfs_log_remove(zilog, tx, txtype, dzp, name);
1818 zfs_dirent_unlock(dl);
1823 /* this rele is delayed to prevent nesting transactions */
1832 * Create a new directory and insert it into dvp using the name
1833 * provided. Return a pointer to the inserted directory.
1835 * IN: dvp - vnode of directory to add subdir to.
1836 * dirname - name of new directory.
1837 * vap - attributes of new directory.
1838 * cr - credentials of caller.
1839 * ct - caller context
1840 * vsecp - ACL to be set
1842 * OUT: vpp - vnode of created directory.
1844 * RETURN: 0 if success
1845 * error code if failure
1848 * dvp - ctime|mtime updated
1849 * vp - ctime|mtime|atime updated
1853 zfs_mkdir(vnode_t *dvp, char *dirname, vattr_t *vap, vnode_t **vpp, cred_t *cr,
1854 caller_context_t *ct, int flags, vsecattr_t *vsecp)
1856 znode_t *zp, *dzp = VTOZ(dvp);
1857 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1866 gid_t gid = crgetgid(cr);
1867 zfs_acl_ids_t acl_ids;
1868 boolean_t fuid_dirtied;
1870 ASSERT(vap->va_type == VDIR);
1873 * If we have an ephemeral id, ACL, or XVATTR then
1874 * make sure file system is at proper version
1877 ksid = crgetsid(cr, KSID_OWNER);
1879 uid = ksid_getid(ksid);
1882 if (zfsvfs->z_use_fuids == B_FALSE &&
1883 (vsecp || (vap->va_mask & AT_XVATTR) || IS_EPHEMERAL(crgetuid(cr))||
1884 IS_EPHEMERAL(crgetgid(cr))))
1889 zilog = zfsvfs->z_log;
1891 if (dzp->z_phys->zp_flags & ZFS_XATTR) {
1896 if (zfsvfs->z_utf8 && u8_validate(dirname,
1897 strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
1901 if (flags & FIGNORECASE)
1904 if (vap->va_mask & AT_XVATTR)
1905 if ((error = secpolicy_xvattr(dvp, (xvattr_t *)vap,
1906 crgetuid(cr), cr, vap->va_type)) != 0) {
1912 * First make sure the new directory doesn't exist.
1917 if (error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf,
1923 if (error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr)) {
1924 zfs_dirent_unlock(dl);
1929 if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, vsecp,
1931 zfs_dirent_unlock(dl);
1935 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
1936 zfs_acl_ids_free(&acl_ids);
1937 zfs_dirent_unlock(dl);
1943 * Add a new entry to the directory.
1945 tx = dmu_tx_create(zfsvfs->z_os);
1946 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname);
1947 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
1948 fuid_dirtied = zfsvfs->z_fuid_dirty;
1950 zfs_fuid_txhold(zfsvfs, tx);
1951 if (acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE)
1952 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
1953 0, SPA_MAXBLOCKSIZE);
1954 error = dmu_tx_assign(tx, TXG_NOWAIT);
1956 zfs_acl_ids_free(&acl_ids);
1957 zfs_dirent_unlock(dl);
1958 if (error == ERESTART) {
1971 zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, &acl_ids);
1974 zfs_fuid_sync(zfsvfs, tx);
1976 * Now put new name in parent dir.
1978 (void) zfs_link_create(dl, zp, tx, ZNEW);
1982 txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap);
1983 if (flags & FIGNORECASE)
1985 zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp,
1986 acl_ids.z_fuidp, vap);
1988 zfs_acl_ids_free(&acl_ids);
1991 zfs_dirent_unlock(dl);
1998 * Remove a directory subdir entry. If the current working
1999 * directory is the same as the subdir to be removed, the
2002 * IN: dvp - vnode of directory to remove from.
2003 * name - name of directory to be removed.
2004 * cwd - vnode of current working directory.
2005 * cr - credentials of caller.
2006 * ct - caller context
2007 * flags - case flags
2009 * RETURN: 0 if success
2010 * error code if failure
2013 * dvp - ctime|mtime updated
2017 zfs_rmdir(vnode_t *dvp, char *name, vnode_t *cwd, cred_t *cr,
2018 caller_context_t *ct, int flags)
2020 znode_t *dzp = VTOZ(dvp);
2023 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
2032 zilog = zfsvfs->z_log;
2034 if (flags & FIGNORECASE)
2040 * Attempt to lock directory; fail if entry doesn't exist.
2042 if (error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
2050 if (error = zfs_zaccess_delete(dzp, zp, cr)) {
2054 if (vp->v_type != VDIR) {
2064 vnevent_rmdir(vp, dvp, name, ct);
2067 * Grab a lock on the directory to make sure that noone is
2068 * trying to add (or lookup) entries while we are removing it.
2070 rw_enter(&zp->z_name_lock, RW_WRITER);
2073 * Grab a lock on the parent pointer to make sure we play well
2074 * with the treewalk and directory rename code.
2076 rw_enter(&zp->z_parent_lock, RW_WRITER);
2078 tx = dmu_tx_create(zfsvfs->z_os);
2079 dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name);
2080 dmu_tx_hold_bonus(tx, zp->z_id);
2081 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
2082 error = dmu_tx_assign(tx, TXG_NOWAIT);
2084 rw_exit(&zp->z_parent_lock);
2085 rw_exit(&zp->z_name_lock);
2086 zfs_dirent_unlock(dl);
2088 if (error == ERESTART) {
2098 #ifdef FREEBSD_NAMECACHE
2102 error = zfs_link_destroy(dl, zp, tx, zflg, NULL);
2105 uint64_t txtype = TX_RMDIR;
2106 if (flags & FIGNORECASE)
2108 zfs_log_remove(zilog, tx, txtype, dzp, name);
2113 rw_exit(&zp->z_parent_lock);
2114 rw_exit(&zp->z_name_lock);
2115 #ifdef FREEBSD_NAMECACHE
2119 zfs_dirent_unlock(dl);
2128 * Read as many directory entries as will fit into the provided
2129 * buffer from the given directory cursor position (specified in
2130 * the uio structure.
2132 * IN: vp - vnode of directory to read.
2133 * uio - structure supplying read location, range info,
2134 * and return buffer.
2135 * cr - credentials of caller.
2136 * ct - caller context
2137 * flags - case flags
2139 * OUT: uio - updated offset and range, buffer filled.
2140 * eofp - set to true if end-of-file detected.
2142 * RETURN: 0 if success
2143 * error code if failure
2146 * vp - atime updated
2148 * Note that the low 4 bits of the cookie returned by zap is always zero.
2149 * This allows us to use the low range for "special" directory entries:
2150 * We use 0 for '.', and 1 for '..'. If this is the root of the filesystem,
2151 * we use the offset 2 for the '.zfs' directory.
2155 zfs_readdir(vnode_t *vp, uio_t *uio, cred_t *cr, int *eofp, int *ncookies, u_long **cookies)
2157 znode_t *zp = VTOZ(vp);
2161 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2166 zap_attribute_t zap;
2167 uint_t bytes_wanted;
2168 uint64_t offset; /* must be unsigned; checks for < 1 */
2173 boolean_t check_sysattrs;
2176 u_long *cooks = NULL;
2183 * If we are not given an eof variable,
2190 * Check for valid iov_len.
2192 if (uio->uio_iov->iov_len <= 0) {
2198 * Quit if directory has been removed (posix)
2200 if ((*eofp = zp->z_unlinked) != 0) {
2207 offset = uio->uio_loffset;
2208 prefetch = zp->z_zn_prefetch;
2211 * Initialize the iterator cursor.
2215 * Start iteration from the beginning of the directory.
2217 zap_cursor_init(&zc, os, zp->z_id);
2220 * The offset is a serialized cursor.
2222 zap_cursor_init_serialized(&zc, os, zp->z_id, offset);
2226 * Get space to change directory entries into fs independent format.
2228 iovp = uio->uio_iov;
2229 bytes_wanted = iovp->iov_len;
2230 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1) {
2231 bufsize = bytes_wanted;
2232 outbuf = kmem_alloc(bufsize, KM_SLEEP);
2233 odp = (struct dirent64 *)outbuf;
2235 bufsize = bytes_wanted;
2236 odp = (struct dirent64 *)iovp->iov_base;
2238 eodp = (struct edirent *)odp;
2240 if (ncookies != NULL) {
2242 * Minimum entry size is dirent size and 1 byte for a file name.
2244 ncooks = uio->uio_resid / (sizeof(struct dirent) - sizeof(((struct dirent *)NULL)->d_name) + 1);
2245 cooks = malloc(ncooks * sizeof(u_long), M_TEMP, M_WAITOK);
2250 * If this VFS supports the system attribute view interface; and
2251 * we're looking at an extended attribute directory; and we care
2252 * about normalization conflicts on this vfs; then we must check
2253 * for normalization conflicts with the sysattr name space.
2256 check_sysattrs = vfs_has_feature(vp->v_vfsp, VFSFT_SYSATTR_VIEWS) &&
2257 (vp->v_flag & V_XATTRDIR) && zfsvfs->z_norm &&
2258 (flags & V_RDDIR_ENTFLAGS);
2264 * Transform to file-system independent format
2267 while (outcount < bytes_wanted) {
2273 * Special case `.', `..', and `.zfs'.
2276 (void) strcpy(zap.za_name, ".");
2277 zap.za_normalization_conflict = 0;
2280 } else if (offset == 1) {
2281 (void) strcpy(zap.za_name, "..");
2282 zap.za_normalization_conflict = 0;
2283 objnum = zp->z_phys->zp_parent;
2285 } else if (offset == 2 && zfs_show_ctldir(zp)) {
2286 (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME);
2287 zap.za_normalization_conflict = 0;
2288 objnum = ZFSCTL_INO_ROOT;
2294 if (error = zap_cursor_retrieve(&zc, &zap)) {
2295 if ((*eofp = (error == ENOENT)) != 0)
2301 if (zap.za_integer_length != 8 ||
2302 zap.za_num_integers != 1) {
2303 cmn_err(CE_WARN, "zap_readdir: bad directory "
2304 "entry, obj = %lld, offset = %lld\n",
2305 (u_longlong_t)zp->z_id,
2306 (u_longlong_t)offset);
2311 objnum = ZFS_DIRENT_OBJ(zap.za_first_integer);
2313 * MacOS X can extract the object type here such as:
2314 * uint8_t type = ZFS_DIRENT_TYPE(zap.za_first_integer);
2316 type = ZFS_DIRENT_TYPE(zap.za_first_integer);
2318 if (check_sysattrs && !zap.za_normalization_conflict) {
2320 zap.za_normalization_conflict =
2321 xattr_sysattr_casechk(zap.za_name);
2323 panic("%s:%u: TODO", __func__, __LINE__);
2328 if (flags & V_RDDIR_ACCFILTER) {
2330 * If we have no access at all, don't include
2331 * this entry in the returned information
2334 if (zfs_zget(zp->z_zfsvfs, objnum, &ezp) != 0)
2336 if (!zfs_has_access(ezp, cr)) {
2343 if (flags & V_RDDIR_ENTFLAGS)
2344 reclen = EDIRENT_RECLEN(strlen(zap.za_name));
2346 reclen = DIRENT64_RECLEN(strlen(zap.za_name));
2349 * Will this entry fit in the buffer?
2351 if (outcount + reclen > bufsize) {
2353 * Did we manage to fit anything in the buffer?
2361 if (flags & V_RDDIR_ENTFLAGS) {
2363 * Add extended flag entry:
2365 eodp->ed_ino = objnum;
2366 eodp->ed_reclen = reclen;
2367 /* NOTE: ed_off is the offset for the *next* entry */
2368 next = &(eodp->ed_off);
2369 eodp->ed_eflags = zap.za_normalization_conflict ?
2370 ED_CASE_CONFLICT : 0;
2371 (void) strncpy(eodp->ed_name, zap.za_name,
2372 EDIRENT_NAMELEN(reclen));
2373 eodp = (edirent_t *)((intptr_t)eodp + reclen);
2378 odp->d_ino = objnum;
2379 odp->d_reclen = reclen;
2380 odp->d_namlen = strlen(zap.za_name);
2381 (void) strlcpy(odp->d_name, zap.za_name, odp->d_namlen + 1);
2383 odp = (dirent64_t *)((intptr_t)odp + reclen);
2387 ASSERT(outcount <= bufsize);
2389 /* Prefetch znode */
2391 dmu_prefetch(os, objnum, 0, 0);
2395 * Move to the next entry, fill in the previous offset.
2397 if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) {
2398 zap_cursor_advance(&zc);
2399 offset = zap_cursor_serialize(&zc);
2404 if (cooks != NULL) {
2407 KASSERT(ncooks >= 0, ("ncookies=%d", ncooks));
2410 zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */
2412 /* Subtract unused cookies */
2413 if (ncookies != NULL)
2414 *ncookies -= ncooks;
2416 if (uio->uio_segflg == UIO_SYSSPACE && uio->uio_iovcnt == 1) {
2417 iovp->iov_base += outcount;
2418 iovp->iov_len -= outcount;
2419 uio->uio_resid -= outcount;
2420 } else if (error = uiomove(outbuf, (long)outcount, UIO_READ, uio)) {
2422 * Reset the pointer.
2424 offset = uio->uio_loffset;
2428 zap_cursor_fini(&zc);
2429 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1)
2430 kmem_free(outbuf, bufsize);
2432 if (error == ENOENT)
2435 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
2437 uio->uio_loffset = offset;
2439 if (error != 0 && cookies != NULL) {
2440 free(*cookies, M_TEMP);
2447 ulong_t zfs_fsync_sync_cnt = 4;
2450 zfs_fsync(vnode_t *vp, int syncflag, cred_t *cr, caller_context_t *ct)
2452 znode_t *zp = VTOZ(vp);
2453 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2455 (void) tsd_set(zfs_fsyncer_key, (void *)zfs_fsync_sync_cnt);
2459 zil_commit(zfsvfs->z_log, zp->z_last_itx, zp->z_id);
2466 * Get the requested file attributes and place them in the provided
2469 * IN: vp - vnode of file.
2470 * vap - va_mask identifies requested attributes.
2471 * If AT_XVATTR set, then optional attrs are requested
2472 * flags - ATTR_NOACLCHECK (CIFS server context)
2473 * cr - credentials of caller.
2474 * ct - caller context
2476 * OUT: vap - attribute values.
2478 * RETURN: 0 (always succeeds)
2482 zfs_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
2483 caller_context_t *ct)
2485 znode_t *zp = VTOZ(vp);
2486 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2490 u_longlong_t nblocks;
2492 xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */
2493 xoptattr_t *xoap = NULL;
2494 boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
2501 * If ACL is trivial don't bother looking for ACE_READ_ATTRIBUTES.
2502 * Also, if we are the owner don't bother, since owner should
2503 * always be allowed to read basic attributes of file.
2505 if (!(pzp->zp_flags & ZFS_ACL_TRIVIAL) &&
2506 (pzp->zp_uid != crgetuid(cr))) {
2507 if (error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0,
2515 * Return all attributes. It's cheaper to provide the answer
2516 * than to determine whether we were asked the question.
2519 mutex_enter(&zp->z_lock);
2520 vap->va_type = IFTOVT(pzp->zp_mode);
2521 vap->va_mode = pzp->zp_mode & ~S_IFMT;
2522 zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid);
2523 // vap->va_fsid = zp->z_zfsvfs->z_vfs->vfs_dev;
2524 vap->va_nodeid = zp->z_id;
2525 if ((vp->v_flag & VROOT) && zfs_show_ctldir(zp))
2526 links = pzp->zp_links + 1;
2528 links = pzp->zp_links;
2529 vap->va_nlink = MIN(links, UINT32_MAX); /* nlink_t limit! */
2530 vap->va_size = pzp->zp_size;
2531 vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
2532 vap->va_rdev = zfs_cmpldev(pzp->zp_rdev);
2533 vap->va_seq = zp->z_seq;
2534 vap->va_flags = 0; /* FreeBSD: Reset chflags(2) flags. */
2537 * Add in any requested optional attributes and the create time.
2538 * Also set the corresponding bits in the returned attribute bitmap.
2540 if ((xoap = xva_getxoptattr(xvap)) != NULL && zfsvfs->z_use_fuids) {
2541 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
2543 ((pzp->zp_flags & ZFS_ARCHIVE) != 0);
2544 XVA_SET_RTN(xvap, XAT_ARCHIVE);
2547 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
2548 xoap->xoa_readonly =
2549 ((pzp->zp_flags & ZFS_READONLY) != 0);
2550 XVA_SET_RTN(xvap, XAT_READONLY);
2553 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
2555 ((pzp->zp_flags & ZFS_SYSTEM) != 0);
2556 XVA_SET_RTN(xvap, XAT_SYSTEM);
2559 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
2561 ((pzp->zp_flags & ZFS_HIDDEN) != 0);
2562 XVA_SET_RTN(xvap, XAT_HIDDEN);
2565 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
2566 xoap->xoa_nounlink =
2567 ((pzp->zp_flags & ZFS_NOUNLINK) != 0);
2568 XVA_SET_RTN(xvap, XAT_NOUNLINK);
2571 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
2572 xoap->xoa_immutable =
2573 ((pzp->zp_flags & ZFS_IMMUTABLE) != 0);
2574 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
2577 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
2578 xoap->xoa_appendonly =
2579 ((pzp->zp_flags & ZFS_APPENDONLY) != 0);
2580 XVA_SET_RTN(xvap, XAT_APPENDONLY);
2583 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
2585 ((pzp->zp_flags & ZFS_NODUMP) != 0);
2586 XVA_SET_RTN(xvap, XAT_NODUMP);
2589 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
2591 ((pzp->zp_flags & ZFS_OPAQUE) != 0);
2592 XVA_SET_RTN(xvap, XAT_OPAQUE);
2595 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
2596 xoap->xoa_av_quarantined =
2597 ((pzp->zp_flags & ZFS_AV_QUARANTINED) != 0);
2598 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
2601 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
2602 xoap->xoa_av_modified =
2603 ((pzp->zp_flags & ZFS_AV_MODIFIED) != 0);
2604 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
2607 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) &&
2608 vp->v_type == VREG &&
2609 (pzp->zp_flags & ZFS_BONUS_SCANSTAMP)) {
2611 dmu_object_info_t doi;
2614 * Only VREG files have anti-virus scanstamps, so we
2615 * won't conflict with symlinks in the bonus buffer.
2617 dmu_object_info_from_db(zp->z_dbuf, &doi);
2618 len = sizeof (xoap->xoa_av_scanstamp) +
2619 sizeof (znode_phys_t);
2620 if (len <= doi.doi_bonus_size) {
2622 * pzp points to the start of the
2623 * znode_phys_t. pzp + 1 points to the
2624 * first byte after the znode_phys_t.
2626 (void) memcpy(xoap->xoa_av_scanstamp,
2628 sizeof (xoap->xoa_av_scanstamp));
2629 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
2633 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
2634 ZFS_TIME_DECODE(&xoap->xoa_createtime, pzp->zp_crtime);
2635 XVA_SET_RTN(xvap, XAT_CREATETIME);
2639 ZFS_TIME_DECODE(&vap->va_atime, pzp->zp_atime);
2640 ZFS_TIME_DECODE(&vap->va_mtime, pzp->zp_mtime);
2641 ZFS_TIME_DECODE(&vap->va_ctime, pzp->zp_ctime);
2642 ZFS_TIME_DECODE(&vap->va_birthtime, pzp->zp_crtime);
2644 mutex_exit(&zp->z_lock);
2646 dmu_object_size_from_db(zp->z_dbuf, &blksize, &nblocks);
2647 vap->va_blksize = blksize;
2648 vap->va_bytes = nblocks << 9; /* nblocks * 512 */
2650 if (zp->z_blksz == 0) {
2652 * Block size hasn't been set; suggest maximal I/O transfers.
2654 vap->va_blksize = zfsvfs->z_max_blksz;
2662 * Set the file attributes to the values contained in the
2665 * IN: vp - vnode of file to be modified.
2666 * vap - new attribute values.
2667 * If AT_XVATTR set, then optional attrs are being set
2668 * flags - ATTR_UTIME set if non-default time values provided.
2669 * - ATTR_NOACLCHECK (CIFS context only).
2670 * cr - credentials of caller.
2671 * ct - caller context
2673 * RETURN: 0 if success
2674 * error code if failure
2677 * vp - ctime updated, mtime updated if size changed.
2681 zfs_setattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
2682 caller_context_t *ct)
2684 znode_t *zp = VTOZ(vp);
2686 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2691 uint_t mask = vap->va_mask;
2693 uint64_t saved_mode;
2696 uint64_t new_uid, new_gid;
2698 int need_policy = FALSE;
2700 zfs_fuid_info_t *fuidp = NULL;
2701 xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */
2703 zfs_acl_t *aclp = NULL;
2704 boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
2705 boolean_t fuid_dirtied = B_FALSE;
2710 if (mask & AT_NOSET)
2717 zilog = zfsvfs->z_log;
2720 * Make sure that if we have ephemeral uid/gid or xvattr specified
2721 * that file system is at proper version level
2724 if (zfsvfs->z_use_fuids == B_FALSE &&
2725 (((mask & AT_UID) && IS_EPHEMERAL(vap->va_uid)) ||
2726 ((mask & AT_GID) && IS_EPHEMERAL(vap->va_gid)) ||
2727 (mask & AT_XVATTR))) {
2732 if (mask & AT_SIZE && vp->v_type == VDIR) {
2737 if (mask & AT_SIZE && vp->v_type != VREG && vp->v_type != VFIFO) {
2743 * If this is an xvattr_t, then get a pointer to the structure of
2744 * optional attributes. If this is NULL, then we have a vattr_t.
2746 xoap = xva_getxoptattr(xvap);
2748 xva_init(&tmpxvattr);
2751 * Immutable files can only alter immutable bit and atime
2753 if ((pzp->zp_flags & ZFS_IMMUTABLE) &&
2754 ((mask & (AT_SIZE|AT_UID|AT_GID|AT_MTIME|AT_MODE)) ||
2755 ((mask & AT_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) {
2760 if ((mask & AT_SIZE) && (pzp->zp_flags & ZFS_READONLY)) {
2766 * Verify timestamps doesn't overflow 32 bits.
2767 * ZFS can handle large timestamps, but 32bit syscalls can't
2768 * handle times greater than 2039. This check should be removed
2769 * once large timestamps are fully supported.
2771 if (mask & (AT_ATIME | AT_MTIME)) {
2772 if (((mask & AT_ATIME) && TIMESPEC_OVERFLOW(&vap->va_atime)) ||
2773 ((mask & AT_MTIME) && TIMESPEC_OVERFLOW(&vap->va_mtime))) {
2782 /* Can this be moved to before the top label? */
2783 if (zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) {
2789 * First validate permissions
2792 if (mask & AT_SIZE) {
2793 err = zfs_zaccess(zp, ACE_WRITE_DATA, 0, skipaclchk, cr);
2799 * XXX - Note, we are not providing any open
2800 * mode flags here (like FNDELAY), so we may
2801 * block if there are locks present... this
2802 * should be addressed in openat().
2804 /* XXX - would it be OK to generate a log record here? */
2805 err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE);
2812 if (mask & (AT_ATIME|AT_MTIME) ||
2813 ((mask & AT_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) ||
2814 XVA_ISSET_REQ(xvap, XAT_READONLY) ||
2815 XVA_ISSET_REQ(xvap, XAT_ARCHIVE) ||
2816 XVA_ISSET_REQ(xvap, XAT_CREATETIME) ||
2817 XVA_ISSET_REQ(xvap, XAT_SYSTEM))))
2818 need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0,
2821 if (mask & (AT_UID|AT_GID)) {
2822 int idmask = (mask & (AT_UID|AT_GID));
2827 * NOTE: even if a new mode is being set,
2828 * we may clear S_ISUID/S_ISGID bits.
2831 if (!(mask & AT_MODE))
2832 vap->va_mode = pzp->zp_mode;
2835 * Take ownership or chgrp to group we are a member of
2838 take_owner = (mask & AT_UID) && (vap->va_uid == crgetuid(cr));
2839 take_group = (mask & AT_GID) &&
2840 zfs_groupmember(zfsvfs, vap->va_gid, cr);
2843 * If both AT_UID and AT_GID are set then take_owner and
2844 * take_group must both be set in order to allow taking
2847 * Otherwise, send the check through secpolicy_vnode_setattr()
2851 if (((idmask == (AT_UID|AT_GID)) && take_owner && take_group) ||
2852 ((idmask == AT_UID) && take_owner) ||
2853 ((idmask == AT_GID) && take_group)) {
2854 if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0,
2855 skipaclchk, cr) == 0) {
2857 * Remove setuid/setgid for non-privileged users
2859 secpolicy_setid_clear(vap, vp, cr);
2860 trim_mask = (mask & (AT_UID|AT_GID));
2869 mutex_enter(&zp->z_lock);
2870 oldva.va_mode = pzp->zp_mode;
2871 zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid);
2872 if (mask & AT_XVATTR) {
2874 * Update xvattr mask to include only those attributes
2875 * that are actually changing.
2877 * the bits will be restored prior to actually setting
2878 * the attributes so the caller thinks they were set.
2880 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
2881 if (xoap->xoa_appendonly !=
2882 ((pzp->zp_flags & ZFS_APPENDONLY) != 0)) {
2885 XVA_CLR_REQ(xvap, XAT_APPENDONLY);
2886 XVA_SET_REQ(&tmpxvattr, XAT_APPENDONLY);
2890 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
2891 if (xoap->xoa_nounlink !=
2892 ((pzp->zp_flags & ZFS_NOUNLINK) != 0)) {
2895 XVA_CLR_REQ(xvap, XAT_NOUNLINK);
2896 XVA_SET_REQ(&tmpxvattr, XAT_NOUNLINK);
2900 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
2901 if (xoap->xoa_immutable !=
2902 ((pzp->zp_flags & ZFS_IMMUTABLE) != 0)) {
2905 XVA_CLR_REQ(xvap, XAT_IMMUTABLE);
2906 XVA_SET_REQ(&tmpxvattr, XAT_IMMUTABLE);
2910 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
2911 if (xoap->xoa_nodump !=
2912 ((pzp->zp_flags & ZFS_NODUMP) != 0)) {
2915 XVA_CLR_REQ(xvap, XAT_NODUMP);
2916 XVA_SET_REQ(&tmpxvattr, XAT_NODUMP);
2920 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
2921 if (xoap->xoa_av_modified !=
2922 ((pzp->zp_flags & ZFS_AV_MODIFIED) != 0)) {
2925 XVA_CLR_REQ(xvap, XAT_AV_MODIFIED);
2926 XVA_SET_REQ(&tmpxvattr, XAT_AV_MODIFIED);
2930 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
2931 if ((vp->v_type != VREG &&
2932 xoap->xoa_av_quarantined) ||
2933 xoap->xoa_av_quarantined !=
2934 ((pzp->zp_flags & ZFS_AV_QUARANTINED) != 0)) {
2937 XVA_CLR_REQ(xvap, XAT_AV_QUARANTINED);
2938 XVA_SET_REQ(&tmpxvattr, XAT_AV_QUARANTINED);
2942 if (need_policy == FALSE &&
2943 (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) ||
2944 XVA_ISSET_REQ(xvap, XAT_OPAQUE))) {
2949 mutex_exit(&zp->z_lock);
2951 if (mask & AT_MODE) {
2952 if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr) == 0) {
2953 err = secpolicy_setid_setsticky_clear(vp, vap,
2959 trim_mask |= AT_MODE;
2967 * If trim_mask is set then take ownership
2968 * has been granted or write_acl is present and user
2969 * has the ability to modify mode. In that case remove
2970 * UID|GID and or MODE from mask so that
2971 * secpolicy_vnode_setattr() doesn't revoke it.
2975 saved_mask = vap->va_mask;
2976 vap->va_mask &= ~trim_mask;
2977 if (trim_mask & AT_MODE) {
2979 * Save the mode, as secpolicy_vnode_setattr()
2980 * will overwrite it with ova.va_mode.
2982 saved_mode = vap->va_mode;
2985 err = secpolicy_vnode_setattr(cr, vp, vap, &oldva, flags,
2986 (int (*)(void *, int, cred_t *))zfs_zaccess_unix, zp);
2993 vap->va_mask |= saved_mask;
2994 if (trim_mask & AT_MODE) {
2996 * Recover the mode after
2997 * secpolicy_vnode_setattr().
2999 vap->va_mode = saved_mode;
3005 * secpolicy_vnode_setattr, or take ownership may have
3008 mask = vap->va_mask;
3010 tx = dmu_tx_create(zfsvfs->z_os);
3011 dmu_tx_hold_bonus(tx, zp->z_id);
3013 if (mask & AT_MODE) {
3014 uint64_t pmode = pzp->zp_mode;
3016 new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT);
3018 if (err = zfs_acl_chmod_setattr(zp, &aclp, new_mode))
3020 if (pzp->zp_acl.z_acl_extern_obj) {
3021 /* Are we upgrading ACL from old V0 format to new V1 */
3022 if (zfsvfs->z_version <= ZPL_VERSION_FUID &&
3023 pzp->zp_acl.z_acl_version ==
3024 ZFS_ACL_VERSION_INITIAL) {
3025 dmu_tx_hold_free(tx,
3026 pzp->zp_acl.z_acl_extern_obj, 0,
3028 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
3029 0, aclp->z_acl_bytes);
3031 dmu_tx_hold_write(tx,
3032 pzp->zp_acl.z_acl_extern_obj, 0,
3035 } else if (aclp->z_acl_bytes > ZFS_ACE_SPACE) {
3036 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
3037 0, aclp->z_acl_bytes);
3041 if (mask & (AT_UID | AT_GID)) {
3042 if (pzp->zp_xattr) {
3043 err = zfs_zget(zp->z_zfsvfs, pzp->zp_xattr, &attrzp);
3046 dmu_tx_hold_bonus(tx, attrzp->z_id);
3048 if (mask & AT_UID) {
3049 new_uid = zfs_fuid_create(zfsvfs,
3050 (uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp);
3051 if (new_uid != pzp->zp_uid &&
3052 zfs_usergroup_overquota(zfsvfs, B_FALSE, new_uid)) {
3058 if (mask & AT_GID) {
3059 new_gid = zfs_fuid_create(zfsvfs, (uint64_t)vap->va_gid,
3060 cr, ZFS_GROUP, &fuidp);
3061 if (new_gid != pzp->zp_gid &&
3062 zfs_usergroup_overquota(zfsvfs, B_TRUE, new_gid)) {
3067 fuid_dirtied = zfsvfs->z_fuid_dirty;
3069 if (zfsvfs->z_fuid_obj == 0) {
3070 dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
3071 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
3072 FUID_SIZE_ESTIMATE(zfsvfs));
3073 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ,
3076 dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj);
3077 dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0,
3078 FUID_SIZE_ESTIMATE(zfsvfs));
3083 err = dmu_tx_assign(tx, TXG_NOWAIT);
3085 if (err == ERESTART)
3090 dmu_buf_will_dirty(zp->z_dbuf, tx);
3093 * Set each attribute requested.
3094 * We group settings according to the locks they need to acquire.
3096 * Note: you cannot set ctime directly, although it will be
3097 * updated as a side-effect of calling this function.
3100 mutex_enter(&zp->z_lock);
3102 if (mask & AT_MODE) {
3103 mutex_enter(&zp->z_acl_lock);
3104 zp->z_phys->zp_mode = new_mode;
3105 err = zfs_aclset_common(zp, aclp, cr, tx);
3106 ASSERT3U(err, ==, 0);
3107 zp->z_acl_cached = aclp;
3109 mutex_exit(&zp->z_acl_lock);
3113 mutex_enter(&attrzp->z_lock);
3115 if (mask & AT_UID) {
3116 pzp->zp_uid = new_uid;
3118 attrzp->z_phys->zp_uid = new_uid;
3121 if (mask & AT_GID) {
3122 pzp->zp_gid = new_gid;
3124 attrzp->z_phys->zp_gid = new_gid;
3128 mutex_exit(&attrzp->z_lock);
3130 if (mask & AT_ATIME)
3131 ZFS_TIME_ENCODE(&vap->va_atime, pzp->zp_atime);
3133 if (mask & AT_MTIME)
3134 ZFS_TIME_ENCODE(&vap->va_mtime, pzp->zp_mtime);
3136 /* XXX - shouldn't this be done *before* the ATIME/MTIME checks? */
3138 zfs_time_stamper_locked(zp, CONTENT_MODIFIED, tx);
3140 zfs_time_stamper_locked(zp, STATE_CHANGED, tx);
3142 * Do this after setting timestamps to prevent timestamp
3143 * update from toggling bit
3146 if (xoap && (mask & AT_XVATTR)) {
3149 * restore trimmed off masks
3150 * so that return masks can be set for caller.
3153 if (XVA_ISSET_REQ(&tmpxvattr, XAT_APPENDONLY)) {
3154 XVA_SET_REQ(xvap, XAT_APPENDONLY);
3156 if (XVA_ISSET_REQ(&tmpxvattr, XAT_NOUNLINK)) {
3157 XVA_SET_REQ(xvap, XAT_NOUNLINK);
3159 if (XVA_ISSET_REQ(&tmpxvattr, XAT_IMMUTABLE)) {
3160 XVA_SET_REQ(xvap, XAT_IMMUTABLE);
3162 if (XVA_ISSET_REQ(&tmpxvattr, XAT_NODUMP)) {
3163 XVA_SET_REQ(xvap, XAT_NODUMP);
3165 if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_MODIFIED)) {
3166 XVA_SET_REQ(xvap, XAT_AV_MODIFIED);
3168 if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_QUARANTINED)) {
3169 XVA_SET_REQ(xvap, XAT_AV_QUARANTINED);
3172 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
3174 dmu_object_info_t doi;
3176 ASSERT(vp->v_type == VREG);
3178 /* Grow the bonus buffer if necessary. */
3179 dmu_object_info_from_db(zp->z_dbuf, &doi);
3180 len = sizeof (xoap->xoa_av_scanstamp) +
3181 sizeof (znode_phys_t);
3182 if (len > doi.doi_bonus_size)
3183 VERIFY(dmu_set_bonus(zp->z_dbuf, len, tx) == 0);
3185 zfs_xvattr_set(zp, xvap);
3189 zfs_fuid_sync(zfsvfs, tx);
3192 zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp);
3194 mutex_exit(&zp->z_lock);
3198 VN_RELE(ZTOV(attrzp));
3204 zfs_fuid_info_free(fuidp);
3213 if (err == ERESTART)
3220 typedef struct zfs_zlock {
3221 krwlock_t *zl_rwlock; /* lock we acquired */
3222 znode_t *zl_znode; /* znode we held */
3223 struct zfs_zlock *zl_next; /* next in list */
3227 * Drop locks and release vnodes that were held by zfs_rename_lock().
3230 zfs_rename_unlock(zfs_zlock_t **zlpp)
3234 while ((zl = *zlpp) != NULL) {
3235 if (zl->zl_znode != NULL)
3236 VN_RELE(ZTOV(zl->zl_znode));
3237 rw_exit(zl->zl_rwlock);
3238 *zlpp = zl->zl_next;
3239 kmem_free(zl, sizeof (*zl));
3244 * Search back through the directory tree, using the ".." entries.
3245 * Lock each directory in the chain to prevent concurrent renames.
3246 * Fail any attempt to move a directory into one of its own descendants.
3247 * XXX - z_parent_lock can overlap with map or grow locks
3250 zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp)
3254 uint64_t rootid = zp->z_zfsvfs->z_root;
3255 uint64_t *oidp = &zp->z_id;
3256 krwlock_t *rwlp = &szp->z_parent_lock;
3257 krw_t rw = RW_WRITER;
3260 * First pass write-locks szp and compares to zp->z_id.
3261 * Later passes read-lock zp and compare to zp->z_parent.
3264 if (!rw_tryenter(rwlp, rw)) {
3266 * Another thread is renaming in this path.
3267 * Note that if we are a WRITER, we don't have any
3268 * parent_locks held yet.
3270 if (rw == RW_READER && zp->z_id > szp->z_id) {
3272 * Drop our locks and restart
3274 zfs_rename_unlock(&zl);
3278 rwlp = &szp->z_parent_lock;
3283 * Wait for other thread to drop its locks
3289 zl = kmem_alloc(sizeof (*zl), KM_SLEEP);
3290 zl->zl_rwlock = rwlp;
3291 zl->zl_znode = NULL;
3292 zl->zl_next = *zlpp;
3295 if (*oidp == szp->z_id) /* We're a descendant of szp */
3298 if (*oidp == rootid) /* We've hit the top */
3301 if (rw == RW_READER) { /* i.e. not the first pass */
3302 int error = zfs_zget(zp->z_zfsvfs, *oidp, &zp);
3307 oidp = &zp->z_phys->zp_parent;
3308 rwlp = &zp->z_parent_lock;
3311 } while (zp->z_id != sdzp->z_id);
3317 * Move an entry from the provided source directory to the target
3318 * directory. Change the entry name as indicated.
3320 * IN: sdvp - Source directory containing the "old entry".
3321 * snm - Old entry name.
3322 * tdvp - Target directory to contain the "new entry".
3323 * tnm - New entry name.
3324 * cr - credentials of caller.
3325 * ct - caller context
3326 * flags - case flags
3328 * RETURN: 0 if success
3329 * error code if failure
3332 * sdvp,tdvp - ctime|mtime updated
3336 zfs_rename(vnode_t *sdvp, char *snm, vnode_t *tdvp, char *tnm, cred_t *cr,
3337 caller_context_t *ct, int flags)
3339 znode_t *tdzp, *szp, *tzp;
3340 znode_t *sdzp = VTOZ(sdvp);
3341 zfsvfs_t *zfsvfs = sdzp->z_zfsvfs;
3344 zfs_dirlock_t *sdl, *tdl;
3347 int cmp, serr, terr;
3352 ZFS_VERIFY_ZP(sdzp);
3353 zilog = zfsvfs->z_log;
3356 * Make sure we have the real vp for the target directory.
3358 if (VOP_REALVP(tdvp, &realvp, ct) == 0)
3361 if (tdvp->v_vfsp != sdvp->v_vfsp || zfsctl_is_node(tdvp)) {
3367 ZFS_VERIFY_ZP(tdzp);
3368 if (zfsvfs->z_utf8 && u8_validate(tnm,
3369 strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3374 if (flags & FIGNORECASE)
3383 * This is to prevent the creation of links into attribute space
3384 * by renaming a linked file into/outof an attribute directory.
3385 * See the comment in zfs_link() for why this is considered bad.
3387 if ((tdzp->z_phys->zp_flags & ZFS_XATTR) !=
3388 (sdzp->z_phys->zp_flags & ZFS_XATTR)) {
3394 * Lock source and target directory entries. To prevent deadlock,
3395 * a lock ordering must be defined. We lock the directory with
3396 * the smallest object id first, or if it's a tie, the one with
3397 * the lexically first name.
3399 if (sdzp->z_id < tdzp->z_id) {
3401 } else if (sdzp->z_id > tdzp->z_id) {
3405 * First compare the two name arguments without
3406 * considering any case folding.
3408 int nofold = (zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER);
3410 cmp = u8_strcmp(snm, tnm, 0, nofold, U8_UNICODE_LATEST, &error);
3411 ASSERT(error == 0 || !zfsvfs->z_utf8);
3414 * POSIX: "If the old argument and the new argument
3415 * both refer to links to the same existing file,
3416 * the rename() function shall return successfully
3417 * and perform no other action."
3423 * If the file system is case-folding, then we may
3424 * have some more checking to do. A case-folding file
3425 * system is either supporting mixed case sensitivity
3426 * access or is completely case-insensitive. Note
3427 * that the file system is always case preserving.
3429 * In mixed sensitivity mode case sensitive behavior
3430 * is the default. FIGNORECASE must be used to
3431 * explicitly request case insensitive behavior.
3433 * If the source and target names provided differ only
3434 * by case (e.g., a request to rename 'tim' to 'Tim'),
3435 * we will treat this as a special case in the
3436 * case-insensitive mode: as long as the source name
3437 * is an exact match, we will allow this to proceed as
3438 * a name-change request.
3440 if ((zfsvfs->z_case == ZFS_CASE_INSENSITIVE ||
3441 (zfsvfs->z_case == ZFS_CASE_MIXED &&
3442 flags & FIGNORECASE)) &&
3443 u8_strcmp(snm, tnm, 0, zfsvfs->z_norm, U8_UNICODE_LATEST,
3446 * case preserving rename request, require exact
3455 * If the source and destination directories are the same, we should
3456 * grab the z_name_lock of that directory only once.
3460 rw_enter(&sdzp->z_name_lock, RW_READER);
3464 serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp,
3465 ZEXISTS | zflg, NULL, NULL);
3466 terr = zfs_dirent_lock(&tdl,
3467 tdzp, tnm, &tzp, ZRENAMING | zflg, NULL, NULL);
3469 terr = zfs_dirent_lock(&tdl,
3470 tdzp, tnm, &tzp, zflg, NULL, NULL);
3471 serr = zfs_dirent_lock(&sdl,
3472 sdzp, snm, &szp, ZEXISTS | ZRENAMING | zflg,
3478 * Source entry invalid or not there.
3481 zfs_dirent_unlock(tdl);
3487 rw_exit(&sdzp->z_name_lock);
3489 if (strcmp(snm, ".") == 0 || strcmp(snm, "..") == 0)
3495 zfs_dirent_unlock(sdl);
3499 rw_exit(&sdzp->z_name_lock);
3501 if (strcmp(tnm, "..") == 0)
3508 * Must have write access at the source to remove the old entry
3509 * and write access at the target to create the new entry.
3510 * Note that if target and source are the same, this can be
3511 * done in a single check.
3514 if (error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr))
3517 if (ZTOV(szp)->v_type == VDIR) {
3519 * Check to make sure rename is valid.
3520 * Can't do a move like this: /usr/a/b to /usr/a/b/c/d
3522 if (error = zfs_rename_lock(szp, tdzp, sdzp, &zl))
3527 * Does target exist?
3531 * Source and target must be the same type.
3533 if (ZTOV(szp)->v_type == VDIR) {
3534 if (ZTOV(tzp)->v_type != VDIR) {
3539 if (ZTOV(tzp)->v_type == VDIR) {
3545 * POSIX dictates that when the source and target
3546 * entries refer to the same file object, rename
3547 * must do nothing and exit without error.
3549 if (szp->z_id == tzp->z_id) {
3555 vnevent_rename_src(ZTOV(szp), sdvp, snm, ct);
3557 vnevent_rename_dest(ZTOV(tzp), tdvp, tnm, ct);
3560 * notify the target directory if it is not the same
3561 * as source directory.
3564 vnevent_rename_dest_dir(tdvp, ct);
3567 tx = dmu_tx_create(zfsvfs->z_os);
3568 dmu_tx_hold_bonus(tx, szp->z_id); /* nlink changes */
3569 dmu_tx_hold_bonus(tx, sdzp->z_id); /* nlink changes */
3570 dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm);
3571 dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm);
3573 dmu_tx_hold_bonus(tx, tdzp->z_id); /* nlink changes */
3575 dmu_tx_hold_bonus(tx, tzp->z_id); /* parent changes */
3576 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
3577 error = dmu_tx_assign(tx, TXG_NOWAIT);
3580 zfs_rename_unlock(&zl);
3581 zfs_dirent_unlock(sdl);
3582 zfs_dirent_unlock(tdl);
3585 rw_exit(&sdzp->z_name_lock);
3590 if (error == ERESTART) {
3600 if (tzp) /* Attempt to remove the existing target */
3601 error = zfs_link_destroy(tdl, tzp, tx, zflg, NULL);
3604 error = zfs_link_create(tdl, szp, tx, ZRENAMING);
3606 szp->z_phys->zp_flags |= ZFS_AV_MODIFIED;
3608 error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL);
3611 zfs_log_rename(zilog, tx,
3612 TX_RENAME | (flags & FIGNORECASE ? TX_CI : 0),
3613 sdzp, sdl->dl_name, tdzp, tdl->dl_name, szp);
3615 /* Update path information for the target vnode */
3616 vn_renamepath(tdvp, ZTOV(szp), tnm, strlen(tnm));
3618 #ifdef FREEBSD_NAMECACHE
3629 zfs_rename_unlock(&zl);
3631 zfs_dirent_unlock(sdl);
3632 zfs_dirent_unlock(tdl);
3635 rw_exit(&sdzp->z_name_lock);
3647 * Insert the indicated symbolic reference entry into the directory.
3649 * IN: dvp - Directory to contain new symbolic link.
3650 * link - Name for new symlink entry.
3651 * vap - Attributes of new entry.
3652 * target - Target path of new symlink.
3653 * cr - credentials of caller.
3654 * ct - caller context
3655 * flags - case flags
3657 * RETURN: 0 if success
3658 * error code if failure
3661 * dvp - ctime|mtime updated
3665 zfs_symlink(vnode_t *dvp, vnode_t **vpp, char *name, vattr_t *vap, char *link,
3666 cred_t *cr, kthread_t *td)
3668 znode_t *zp, *dzp = VTOZ(dvp);
3671 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
3673 int len = strlen(link);
3676 zfs_acl_ids_t acl_ids;
3677 boolean_t fuid_dirtied;
3680 ASSERT(vap->va_type == VLNK);
3684 zilog = zfsvfs->z_log;
3686 if (zfsvfs->z_utf8 && u8_validate(name, strlen(name),
3687 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3691 if (flags & FIGNORECASE)
3694 if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) {
3699 if (len > MAXPATHLEN) {
3701 return (ENAMETOOLONG);
3705 * Attempt to lock directory; fail if entry already exists.
3707 error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL);
3713 VERIFY(0 == zfs_acl_ids_create(dzp, 0, vap, cr, NULL, &acl_ids));
3714 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
3715 zfs_acl_ids_free(&acl_ids);
3716 zfs_dirent_unlock(dl);
3720 tx = dmu_tx_create(zfsvfs->z_os);
3721 fuid_dirtied = zfsvfs->z_fuid_dirty;
3722 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len));
3723 dmu_tx_hold_bonus(tx, dzp->z_id);
3724 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
3725 if (acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE)
3726 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, SPA_MAXBLOCKSIZE);
3728 zfs_fuid_txhold(zfsvfs, tx);
3729 error = dmu_tx_assign(tx, TXG_NOWAIT);
3731 zfs_acl_ids_free(&acl_ids);
3732 zfs_dirent_unlock(dl);
3733 if (error == ERESTART) {
3743 dmu_buf_will_dirty(dzp->z_dbuf, tx);
3746 * Create a new object for the symlink.
3747 * Put the link content into bonus buffer if it will fit;
3748 * otherwise, store it just like any other file data.
3750 if (sizeof (znode_phys_t) + len <= dmu_bonus_max()) {
3751 zfs_mknode(dzp, vap, tx, cr, 0, &zp, len, &acl_ids);
3753 bcopy(link, zp->z_phys + 1, len);
3757 zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, &acl_ids);
3760 zfs_fuid_sync(zfsvfs, tx);
3762 * Nothing can access the znode yet so no locking needed
3763 * for growing the znode's blocksize.
3765 zfs_grow_blocksize(zp, len, tx);
3767 VERIFY(0 == dmu_buf_hold(zfsvfs->z_os,
3768 zp->z_id, 0, FTAG, &dbp));
3769 dmu_buf_will_dirty(dbp, tx);
3771 ASSERT3U(len, <=, dbp->db_size);
3772 bcopy(link, dbp->db_data, len);
3773 dmu_buf_rele(dbp, FTAG);
3775 zp->z_phys->zp_size = len;
3778 * Insert the new object into the directory.
3780 (void) zfs_link_create(dl, zp, tx, ZNEW);
3782 uint64_t txtype = TX_SYMLINK;
3783 if (flags & FIGNORECASE)
3785 zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link);
3789 zfs_acl_ids_free(&acl_ids);
3793 zfs_dirent_unlock(dl);
3800 * Return, in the buffer contained in the provided uio structure,
3801 * the symbolic path referred to by vp.
3803 * IN: vp - vnode of symbolic link.
3804 * uoip - structure to contain the link path.
3805 * cr - credentials of caller.
3806 * ct - caller context
3808 * OUT: uio - structure to contain the link path.
3810 * RETURN: 0 if success
3811 * error code if failure
3814 * vp - atime updated
3818 zfs_readlink(vnode_t *vp, uio_t *uio, cred_t *cr, caller_context_t *ct)
3820 znode_t *zp = VTOZ(vp);
3821 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
3828 bufsz = (size_t)zp->z_phys->zp_size;
3829 if (bufsz + sizeof (znode_phys_t) <= zp->z_dbuf->db_size) {
3830 error = uiomove(zp->z_phys + 1,
3831 MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio);
3834 error = dmu_buf_hold(zfsvfs->z_os, zp->z_id, 0, FTAG, &dbp);
3839 error = uiomove(dbp->db_data,
3840 MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio);
3841 dmu_buf_rele(dbp, FTAG);
3844 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
3850 * Insert a new entry into directory tdvp referencing svp.
3852 * IN: tdvp - Directory to contain new entry.
3853 * svp - vnode of new entry.
3854 * name - name of new entry.
3855 * cr - credentials of caller.
3856 * ct - caller context
3858 * RETURN: 0 if success
3859 * error code if failure
3862 * tdvp - ctime|mtime updated
3863 * svp - ctime updated
3867 zfs_link(vnode_t *tdvp, vnode_t *svp, char *name, cred_t *cr,
3868 caller_context_t *ct, int flags)
3870 znode_t *dzp = VTOZ(tdvp);
3872 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
3882 ASSERT(tdvp->v_type == VDIR);
3886 zilog = zfsvfs->z_log;
3888 if (VOP_REALVP(svp, &realvp, ct) == 0)
3892 * POSIX dictates that we return EPERM here.
3893 * Better choices include ENOTSUP or EISDIR.
3895 if (svp->v_type == VDIR) {
3900 if (svp->v_vfsp != tdvp->v_vfsp || zfsctl_is_node(svp)) {
3908 /* Prevent links to .zfs/shares files */
3910 if (szp->z_phys->zp_parent == zfsvfs->z_shares_dir) {
3915 if (zfsvfs->z_utf8 && u8_validate(name,
3916 strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3920 if (flags & FIGNORECASE)
3924 * We do not support links between attributes and non-attributes
3925 * because of the potential security risk of creating links
3926 * into "normal" file space in order to circumvent restrictions
3927 * imposed in attribute space.
3929 if ((szp->z_phys->zp_flags & ZFS_XATTR) !=
3930 (dzp->z_phys->zp_flags & ZFS_XATTR)) {
3936 owner = zfs_fuid_map_id(zfsvfs, szp->z_phys->zp_uid, cr, ZFS_OWNER);
3937 if (owner != crgetuid(cr) &&
3938 secpolicy_basic_link(svp, cr) != 0) {
3943 if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) {
3950 * Attempt to lock directory; fail if entry already exists.
3952 error = zfs_dirent_lock(&dl, dzp, name, &tzp, zf, NULL, NULL);
3958 tx = dmu_tx_create(zfsvfs->z_os);
3959 dmu_tx_hold_bonus(tx, szp->z_id);
3960 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
3961 error = dmu_tx_assign(tx, TXG_NOWAIT);
3963 zfs_dirent_unlock(dl);
3964 if (error == ERESTART) {
3974 error = zfs_link_create(dl, szp, tx, 0);
3977 uint64_t txtype = TX_LINK;
3978 if (flags & FIGNORECASE)
3980 zfs_log_link(zilog, tx, txtype, dzp, szp, name);
3985 zfs_dirent_unlock(dl);
3988 vnevent_link(svp, ct);
3997 zfs_inactive(vnode_t *vp, cred_t *cr, caller_context_t *ct)
3999 znode_t *zp = VTOZ(vp);
4000 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4003 rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER);
4004 if (zp->z_dbuf == NULL) {
4006 * The fs has been unmounted, or we did a
4007 * suspend/resume and this file no longer exists.
4010 vp->v_count = 0; /* count arrives as 1 */
4012 vrecycle(vp, curthread);
4013 rw_exit(&zfsvfs->z_teardown_inactive_lock);
4017 if (zp->z_atime_dirty && zp->z_unlinked == 0) {
4018 dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os);
4020 dmu_tx_hold_bonus(tx, zp->z_id);
4021 error = dmu_tx_assign(tx, TXG_WAIT);
4025 dmu_buf_will_dirty(zp->z_dbuf, tx);
4026 mutex_enter(&zp->z_lock);
4027 zp->z_atime_dirty = 0;
4028 mutex_exit(&zp->z_lock);
4034 rw_exit(&zfsvfs->z_teardown_inactive_lock);
4037 CTASSERT(sizeof(struct zfid_short) <= sizeof(struct fid));
4038 CTASSERT(sizeof(struct zfid_long) <= sizeof(struct fid));
4042 zfs_fid(vnode_t *vp, fid_t *fidp, caller_context_t *ct)
4044 znode_t *zp = VTOZ(vp);
4045 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4047 uint64_t object = zp->z_id;
4053 gen = (uint32_t)zp->z_gen;
4055 size = (zfsvfs->z_parent != zfsvfs) ? LONG_FID_LEN : SHORT_FID_LEN;
4056 fidp->fid_len = size;
4058 zfid = (zfid_short_t *)fidp;
4060 zfid->zf_len = size;
4062 for (i = 0; i < sizeof (zfid->zf_object); i++)
4063 zfid->zf_object[i] = (uint8_t)(object >> (8 * i));
4065 /* Must have a non-zero generation number to distinguish from .zfs */
4068 for (i = 0; i < sizeof (zfid->zf_gen); i++)
4069 zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i));
4071 if (size == LONG_FID_LEN) {
4072 uint64_t objsetid = dmu_objset_id(zfsvfs->z_os);
4075 zlfid = (zfid_long_t *)fidp;
4077 for (i = 0; i < sizeof (zlfid->zf_setid); i++)
4078 zlfid->zf_setid[i] = (uint8_t)(objsetid >> (8 * i));
4080 /* XXX - this should be the generation number for the objset */
4081 for (i = 0; i < sizeof (zlfid->zf_setgen); i++)
4082 zlfid->zf_setgen[i] = 0;
4090 zfs_pathconf(vnode_t *vp, int cmd, ulong_t *valp, cred_t *cr,
4091 caller_context_t *ct)
4103 case _PC_FILESIZEBITS:
4108 case _PC_XATTR_EXISTS:
4110 zfsvfs = zp->z_zfsvfs;
4114 error = zfs_dirent_lock(&dl, zp, "", &xzp,
4115 ZXATTR | ZEXISTS | ZSHARED, NULL, NULL);
4117 zfs_dirent_unlock(dl);
4118 if (!zfs_dirempty(xzp))
4121 } else if (error == ENOENT) {
4123 * If there aren't extended attributes, it's the
4124 * same as having zero of them.
4132 case _PC_ACL_EXTENDED:
4140 case _PC_ACL_PATH_MAX:
4141 *valp = ACL_MAX_ENTRIES;
4144 case _PC_MIN_HOLE_SIZE:
4145 *valp = (int)SPA_MINBLOCKSIZE;
4149 return (EOPNOTSUPP);
4155 zfs_getsecattr(vnode_t *vp, vsecattr_t *vsecp, int flag, cred_t *cr,
4156 caller_context_t *ct)
4158 znode_t *zp = VTOZ(vp);
4159 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4161 boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
4165 error = zfs_getacl(zp, vsecp, skipaclchk, cr);
4173 zfs_setsecattr(vnode_t *vp, vsecattr_t *vsecp, int flag, cred_t *cr,
4174 caller_context_t *ct)
4176 znode_t *zp = VTOZ(vp);
4177 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4179 boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
4183 error = zfs_setacl(zp, vsecp, skipaclchk, cr);
4189 zfs_freebsd_open(ap)
4190 struct vop_open_args /* {
4193 struct ucred *a_cred;
4194 struct thread *a_td;
4197 vnode_t *vp = ap->a_vp;
4198 znode_t *zp = VTOZ(vp);
4201 error = zfs_open(&vp, ap->a_mode, ap->a_cred, NULL);
4203 vnode_create_vobject(vp, zp->z_phys->zp_size, ap->a_td);
4208 zfs_freebsd_close(ap)
4209 struct vop_close_args /* {
4212 struct ucred *a_cred;
4213 struct thread *a_td;
4217 return (zfs_close(ap->a_vp, ap->a_fflag, 0, 0, ap->a_cred, NULL));
4221 zfs_freebsd_ioctl(ap)
4222 struct vop_ioctl_args /* {
4232 return (zfs_ioctl(ap->a_vp, ap->a_command, (intptr_t)ap->a_data,
4233 ap->a_fflag, ap->a_cred, NULL, NULL));
4237 zfs_freebsd_read(ap)
4238 struct vop_read_args /* {
4242 struct ucred *a_cred;
4246 return (zfs_read(ap->a_vp, ap->a_uio, ap->a_ioflag, ap->a_cred, NULL));
4250 zfs_freebsd_write(ap)
4251 struct vop_write_args /* {
4255 struct ucred *a_cred;
4259 return (zfs_write(ap->a_vp, ap->a_uio, ap->a_ioflag, ap->a_cred, NULL));
4263 zfs_freebsd_access(ap)
4264 struct vop_access_args /* {
4266 accmode_t a_accmode;
4267 struct ucred *a_cred;
4268 struct thread *a_td;
4275 * ZFS itself only knowns about VREAD, VWRITE, VEXEC and VAPPEND,
4277 accmode = ap->a_accmode & (VREAD|VWRITE|VEXEC|VAPPEND);
4279 error = zfs_access(ap->a_vp, accmode, 0, ap->a_cred, NULL);
4282 * VADMIN has to be handled by vaccess().
4285 accmode = ap->a_accmode & ~(VREAD|VWRITE|VEXEC|VAPPEND);
4287 vnode_t *vp = ap->a_vp;
4288 znode_t *zp = VTOZ(vp);
4289 znode_phys_t *zphys = zp->z_phys;
4291 error = vaccess(vp->v_type, zphys->zp_mode,
4292 zphys->zp_uid, zphys->zp_gid, accmode, ap->a_cred,
4301 zfs_freebsd_lookup(ap)
4302 struct vop_lookup_args /* {
4303 struct vnode *a_dvp;
4304 struct vnode **a_vpp;
4305 struct componentname *a_cnp;
4308 struct componentname *cnp = ap->a_cnp;
4309 char nm[NAME_MAX + 1];
4311 ASSERT(cnp->cn_namelen < sizeof(nm));
4312 strlcpy(nm, cnp->cn_nameptr, MIN(cnp->cn_namelen + 1, sizeof(nm)));
4314 return (zfs_lookup(ap->a_dvp, nm, ap->a_vpp, cnp, cnp->cn_nameiop,
4315 cnp->cn_cred, cnp->cn_thread, 0));
4319 zfs_freebsd_create(ap)
4320 struct vop_create_args /* {
4321 struct vnode *a_dvp;
4322 struct vnode **a_vpp;
4323 struct componentname *a_cnp;
4324 struct vattr *a_vap;
4327 struct componentname *cnp = ap->a_cnp;
4328 vattr_t *vap = ap->a_vap;
4331 ASSERT(cnp->cn_flags & SAVENAME);
4333 vattr_init_mask(vap);
4334 mode = vap->va_mode & ALLPERMS;
4336 return (zfs_create(ap->a_dvp, cnp->cn_nameptr, vap, !EXCL, mode,
4337 ap->a_vpp, cnp->cn_cred, cnp->cn_thread));
4341 zfs_freebsd_remove(ap)
4342 struct vop_remove_args /* {
4343 struct vnode *a_dvp;
4345 struct componentname *a_cnp;
4349 ASSERT(ap->a_cnp->cn_flags & SAVENAME);
4351 return (zfs_remove(ap->a_dvp, ap->a_cnp->cn_nameptr,
4352 ap->a_cnp->cn_cred, NULL, 0));
4356 zfs_freebsd_mkdir(ap)
4357 struct vop_mkdir_args /* {
4358 struct vnode *a_dvp;
4359 struct vnode **a_vpp;
4360 struct componentname *a_cnp;
4361 struct vattr *a_vap;
4364 vattr_t *vap = ap->a_vap;
4366 ASSERT(ap->a_cnp->cn_flags & SAVENAME);
4368 vattr_init_mask(vap);
4370 return (zfs_mkdir(ap->a_dvp, ap->a_cnp->cn_nameptr, vap, ap->a_vpp,
4371 ap->a_cnp->cn_cred, NULL, 0, NULL));
4375 zfs_freebsd_rmdir(ap)
4376 struct vop_rmdir_args /* {
4377 struct vnode *a_dvp;
4379 struct componentname *a_cnp;
4382 struct componentname *cnp = ap->a_cnp;
4384 ASSERT(cnp->cn_flags & SAVENAME);
4386 return (zfs_rmdir(ap->a_dvp, cnp->cn_nameptr, NULL, cnp->cn_cred, NULL, 0));
4390 zfs_freebsd_readdir(ap)
4391 struct vop_readdir_args /* {
4394 struct ucred *a_cred;
4401 return (zfs_readdir(ap->a_vp, ap->a_uio, ap->a_cred, ap->a_eofflag,
4402 ap->a_ncookies, ap->a_cookies));
4406 zfs_freebsd_fsync(ap)
4407 struct vop_fsync_args /* {
4410 struct thread *a_td;
4415 return (zfs_fsync(ap->a_vp, 0, ap->a_td->td_ucred, NULL));
4419 zfs_freebsd_getattr(ap)
4420 struct vop_getattr_args /* {
4422 struct vattr *a_vap;
4423 struct ucred *a_cred;
4424 struct thread *a_td;
4427 vattr_t *vap = ap->a_vap;
4433 xvap.xva_vattr = *vap;
4434 xvap.xva_vattr.va_mask |= AT_XVATTR;
4436 /* Convert chflags into ZFS-type flags. */
4437 /* XXX: what about SF_SETTABLE?. */
4438 XVA_SET_REQ(&xvap, XAT_IMMUTABLE);
4439 XVA_SET_REQ(&xvap, XAT_APPENDONLY);
4440 XVA_SET_REQ(&xvap, XAT_NOUNLINK);
4441 XVA_SET_REQ(&xvap, XAT_NODUMP);
4442 error = zfs_getattr(ap->a_vp, (vattr_t *)&xvap, 0, ap->a_cred, NULL);
4446 /* Convert ZFS xattr into chflags. */
4447 #define FLAG_CHECK(fflag, xflag, xfield) do { \
4448 if (XVA_ISSET_RTN(&xvap, (xflag)) && (xfield) != 0) \
4449 fflags |= (fflag); \
4451 FLAG_CHECK(SF_IMMUTABLE, XAT_IMMUTABLE,
4452 xvap.xva_xoptattrs.xoa_immutable);
4453 FLAG_CHECK(SF_APPEND, XAT_APPENDONLY,
4454 xvap.xva_xoptattrs.xoa_appendonly);
4455 FLAG_CHECK(SF_NOUNLINK, XAT_NOUNLINK,
4456 xvap.xva_xoptattrs.xoa_nounlink);
4457 FLAG_CHECK(UF_NODUMP, XAT_NODUMP,
4458 xvap.xva_xoptattrs.xoa_nodump);
4460 *vap = xvap.xva_vattr;
4461 vap->va_flags = fflags;
4466 zfs_freebsd_setattr(ap)
4467 struct vop_setattr_args /* {
4469 struct vattr *a_vap;
4470 struct ucred *a_cred;
4471 struct thread *a_td;
4474 vnode_t *vp = ap->a_vp;
4475 vattr_t *vap = ap->a_vap;
4476 cred_t *cred = ap->a_cred;
4481 vattr_init_mask(vap);
4482 vap->va_mask &= ~AT_NOSET;
4485 xvap.xva_vattr = *vap;
4487 zflags = VTOZ(vp)->z_phys->zp_flags;
4489 if (vap->va_flags != VNOVAL) {
4490 zfsvfs_t *zfsvfs = VTOZ(vp)->z_zfsvfs;
4493 if (zfsvfs->z_use_fuids == B_FALSE)
4494 return (EOPNOTSUPP);
4496 fflags = vap->va_flags;
4497 if ((fflags & ~(SF_IMMUTABLE|SF_APPEND|SF_NOUNLINK|UF_NODUMP)) != 0)
4498 return (EOPNOTSUPP);
4500 * Unprivileged processes are not permitted to unset system
4501 * flags, or modify flags if any system flags are set.
4502 * Privileged non-jail processes may not modify system flags
4503 * if securelevel > 0 and any existing system flags are set.
4504 * Privileged jail processes behave like privileged non-jail
4505 * processes if the security.jail.chflags_allowed sysctl is
4506 * is non-zero; otherwise, they behave like unprivileged
4509 if (secpolicy_fs_owner(vp->v_mount, cred) == 0 ||
4510 priv_check_cred(cred, PRIV_VFS_SYSFLAGS, 0) == 0) {
4512 (ZFS_IMMUTABLE | ZFS_APPENDONLY | ZFS_NOUNLINK)) {
4513 error = securelevel_gt(cred, 0);
4519 * Callers may only modify the file flags on objects they
4520 * have VADMIN rights for.
4522 if ((error = VOP_ACCESS(vp, VADMIN, cred, curthread)) != 0)
4525 (ZFS_IMMUTABLE | ZFS_APPENDONLY | ZFS_NOUNLINK)) {
4529 (SF_IMMUTABLE | SF_APPEND | SF_NOUNLINK)) {
4534 #define FLAG_CHANGE(fflag, zflag, xflag, xfield) do { \
4535 if (((fflags & (fflag)) && !(zflags & (zflag))) || \
4536 ((zflags & (zflag)) && !(fflags & (fflag)))) { \
4537 XVA_SET_REQ(&xvap, (xflag)); \
4538 (xfield) = ((fflags & (fflag)) != 0); \
4541 /* Convert chflags into ZFS-type flags. */
4542 /* XXX: what about SF_SETTABLE?. */
4543 FLAG_CHANGE(SF_IMMUTABLE, ZFS_IMMUTABLE, XAT_IMMUTABLE,
4544 xvap.xva_xoptattrs.xoa_immutable);
4545 FLAG_CHANGE(SF_APPEND, ZFS_APPENDONLY, XAT_APPENDONLY,
4546 xvap.xva_xoptattrs.xoa_appendonly);
4547 FLAG_CHANGE(SF_NOUNLINK, ZFS_NOUNLINK, XAT_NOUNLINK,
4548 xvap.xva_xoptattrs.xoa_nounlink);
4549 FLAG_CHANGE(UF_NODUMP, ZFS_NODUMP, XAT_NODUMP,
4550 xvap.xva_xoptattrs.xoa_nodump);
4553 return (zfs_setattr(vp, (vattr_t *)&xvap, 0, cred, NULL));
4557 zfs_freebsd_rename(ap)
4558 struct vop_rename_args /* {
4559 struct vnode *a_fdvp;
4560 struct vnode *a_fvp;
4561 struct componentname *a_fcnp;
4562 struct vnode *a_tdvp;
4563 struct vnode *a_tvp;
4564 struct componentname *a_tcnp;
4567 vnode_t *fdvp = ap->a_fdvp;
4568 vnode_t *fvp = ap->a_fvp;
4569 vnode_t *tdvp = ap->a_tdvp;
4570 vnode_t *tvp = ap->a_tvp;
4573 ASSERT(ap->a_fcnp->cn_flags & (SAVENAME|SAVESTART));
4574 ASSERT(ap->a_tcnp->cn_flags & (SAVENAME|SAVESTART));
4576 error = zfs_rename(fdvp, ap->a_fcnp->cn_nameptr, tdvp,
4577 ap->a_tcnp->cn_nameptr, ap->a_fcnp->cn_cred, NULL, 0);
4592 zfs_freebsd_symlink(ap)
4593 struct vop_symlink_args /* {
4594 struct vnode *a_dvp;
4595 struct vnode **a_vpp;
4596 struct componentname *a_cnp;
4597 struct vattr *a_vap;
4601 struct componentname *cnp = ap->a_cnp;
4602 vattr_t *vap = ap->a_vap;
4604 ASSERT(cnp->cn_flags & SAVENAME);
4606 vap->va_type = VLNK; /* FreeBSD: Syscall only sets va_mode. */
4607 vattr_init_mask(vap);
4609 return (zfs_symlink(ap->a_dvp, ap->a_vpp, cnp->cn_nameptr, vap,
4610 ap->a_target, cnp->cn_cred, cnp->cn_thread));
4614 zfs_freebsd_readlink(ap)
4615 struct vop_readlink_args /* {
4618 struct ucred *a_cred;
4622 return (zfs_readlink(ap->a_vp, ap->a_uio, ap->a_cred, NULL));
4626 zfs_freebsd_link(ap)
4627 struct vop_link_args /* {
4628 struct vnode *a_tdvp;
4630 struct componentname *a_cnp;
4633 struct componentname *cnp = ap->a_cnp;
4635 ASSERT(cnp->cn_flags & SAVENAME);
4637 return (zfs_link(ap->a_tdvp, ap->a_vp, cnp->cn_nameptr, cnp->cn_cred, NULL, 0));
4641 zfs_freebsd_inactive(ap)
4642 struct vop_inactive_args /* {
4644 struct thread *a_td;
4647 vnode_t *vp = ap->a_vp;
4649 zfs_inactive(vp, ap->a_td->td_ucred, NULL);
4654 zfs_reclaim_complete(void *arg, int pending)
4657 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4659 rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER);
4660 if (zp->z_dbuf != NULL) {
4661 ZFS_OBJ_HOLD_ENTER(zfsvfs, zp->z_id);
4662 zfs_znode_dmu_fini(zp);
4663 ZFS_OBJ_HOLD_EXIT(zfsvfs, zp->z_id);
4666 rw_exit(&zfsvfs->z_teardown_inactive_lock);
4668 * If the file system is being unmounted, there is a process waiting
4669 * for us, wake it up.
4671 if (zfsvfs->z_unmounted)
4676 zfs_freebsd_reclaim(ap)
4677 struct vop_reclaim_args /* {
4679 struct thread *a_td;
4682 vnode_t *vp = ap->a_vp;
4683 znode_t *zp = VTOZ(vp);
4684 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4686 rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER);
4691 * Destroy the vm object and flush associated pages.
4693 vnode_destroy_vobject(vp);
4695 mutex_enter(&zp->z_lock);
4696 ASSERT(zp->z_phys != NULL);
4698 mutex_exit(&zp->z_lock);
4702 else if (zp->z_dbuf == NULL)
4704 else /* if (!zp->z_unlinked && zp->z_dbuf != NULL) */ {
4707 locked = MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id)) ? 2 :
4708 ZFS_OBJ_HOLD_TRYENTER(zfsvfs, zp->z_id);
4711 * Lock can't be obtained due to deadlock possibility,
4712 * so defer znode destruction.
4714 TASK_INIT(&zp->z_task, 0, zfs_reclaim_complete, zp);
4715 taskqueue_enqueue(taskqueue_thread, &zp->z_task);
4717 zfs_znode_dmu_fini(zp);
4719 ZFS_OBJ_HOLD_EXIT(zfsvfs, zp->z_id);
4725 ASSERT(vp->v_holdcnt >= 1);
4727 rw_exit(&zfsvfs->z_teardown_inactive_lock);
4733 struct vop_fid_args /* {
4739 return (zfs_fid(ap->a_vp, (void *)ap->a_fid, NULL));
4743 zfs_freebsd_pathconf(ap)
4744 struct vop_pathconf_args /* {
4747 register_t *a_retval;
4753 error = zfs_pathconf(ap->a_vp, ap->a_name, &val, curthread->td_ucred, NULL);
4755 *ap->a_retval = val;
4756 else if (error == EOPNOTSUPP)
4757 error = vop_stdpathconf(ap);
4762 zfs_freebsd_fifo_pathconf(ap)
4763 struct vop_pathconf_args /* {
4766 register_t *a_retval;
4770 switch (ap->a_name) {
4771 case _PC_ACL_EXTENDED:
4773 case _PC_ACL_PATH_MAX:
4774 case _PC_MAC_PRESENT:
4775 return (zfs_freebsd_pathconf(ap));
4777 return (fifo_specops.vop_pathconf(ap));
4782 * FreeBSD's extended attributes namespace defines file name prefix for ZFS'
4783 * extended attribute name:
4786 * system freebsd:system:
4787 * user (none, can be used to access ZFS fsattr(5) attributes
4788 * created on Solaris)
4791 zfs_create_attrname(int attrnamespace, const char *name, char *attrname,
4794 const char *namespace, *prefix, *suffix;
4796 /* We don't allow '/' character in attribute name. */
4797 if (strchr(name, '/') != NULL)
4799 /* We don't allow attribute names that start with "freebsd:" string. */
4800 if (strncmp(name, "freebsd:", 8) == 0)
4803 bzero(attrname, size);
4805 switch (attrnamespace) {
4806 case EXTATTR_NAMESPACE_USER:
4808 prefix = "freebsd:";
4809 namespace = EXTATTR_NAMESPACE_USER_STRING;
4813 * This is the default namespace by which we can access all
4814 * attributes created on Solaris.
4816 prefix = namespace = suffix = "";
4819 case EXTATTR_NAMESPACE_SYSTEM:
4820 prefix = "freebsd:";
4821 namespace = EXTATTR_NAMESPACE_SYSTEM_STRING;
4824 case EXTATTR_NAMESPACE_EMPTY:
4828 if (snprintf(attrname, size, "%s%s%s%s", prefix, namespace, suffix,
4830 return (ENAMETOOLONG);
4836 * Vnode operating to retrieve a named extended attribute.
4839 zfs_getextattr(struct vop_getextattr_args *ap)
4842 IN struct vnode *a_vp;
4843 IN int a_attrnamespace;
4844 IN const char *a_name;
4845 INOUT struct uio *a_uio;
4847 IN struct ucred *a_cred;
4848 IN struct thread *a_td;
4852 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
4853 struct thread *td = ap->a_td;
4854 struct nameidata nd;
4857 vnode_t *xvp = NULL, *vp;
4860 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
4861 ap->a_cred, ap->a_td, VREAD);
4865 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
4872 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
4880 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | MPSAFE, UIO_SYSSPACE, attrname,
4882 error = vn_open_cred(&nd, &flags, 0, 0, ap->a_cred, NULL);
4884 NDFREE(&nd, NDF_ONLY_PNBUF);
4887 if (error == ENOENT)
4892 if (ap->a_size != NULL) {
4893 error = VOP_GETATTR(vp, &va, ap->a_cred);
4895 *ap->a_size = (size_t)va.va_size;
4896 } else if (ap->a_uio != NULL)
4897 error = VOP_READ(vp, ap->a_uio, IO_UNIT | IO_SYNC, ap->a_cred);
4900 vn_close(vp, flags, ap->a_cred, td);
4907 * Vnode operation to remove a named attribute.
4910 zfs_deleteextattr(struct vop_deleteextattr_args *ap)
4913 IN struct vnode *a_vp;
4914 IN int a_attrnamespace;
4915 IN const char *a_name;
4916 IN struct ucred *a_cred;
4917 IN struct thread *a_td;
4921 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
4922 struct thread *td = ap->a_td;
4923 struct nameidata nd;
4926 vnode_t *xvp = NULL, *vp;
4929 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
4930 ap->a_cred, ap->a_td, VWRITE);
4934 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
4941 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
4948 NDINIT_ATVP(&nd, DELETE, NOFOLLOW | LOCKPARENT | LOCKLEAF | MPSAFE,
4949 UIO_SYSSPACE, attrname, xvp, td);
4952 NDFREE(&nd, NDF_ONLY_PNBUF);
4955 if (error == ENOENT)
4959 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
4962 if (vp == nd.ni_dvp)
4972 * Vnode operation to set a named attribute.
4975 zfs_setextattr(struct vop_setextattr_args *ap)
4978 IN struct vnode *a_vp;
4979 IN int a_attrnamespace;
4980 IN const char *a_name;
4981 INOUT struct uio *a_uio;
4982 IN struct ucred *a_cred;
4983 IN struct thread *a_td;
4987 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
4988 struct thread *td = ap->a_td;
4989 struct nameidata nd;
4992 vnode_t *xvp = NULL, *vp;
4995 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
4996 ap->a_cred, ap->a_td, VWRITE);
5000 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
5007 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
5008 LOOKUP_XATTR | CREATE_XATTR_DIR);
5014 flags = FFLAGS(O_WRONLY | O_CREAT);
5015 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | MPSAFE, UIO_SYSSPACE, attrname,
5017 error = vn_open_cred(&nd, &flags, 0600, 0, ap->a_cred, NULL);
5019 NDFREE(&nd, NDF_ONLY_PNBUF);
5027 error = VOP_SETATTR(vp, &va, ap->a_cred);
5029 VOP_WRITE(vp, ap->a_uio, IO_UNIT | IO_SYNC, ap->a_cred);
5032 vn_close(vp, flags, ap->a_cred, td);
5039 * Vnode operation to retrieve extended attributes on a vnode.
5042 zfs_listextattr(struct vop_listextattr_args *ap)
5045 IN struct vnode *a_vp;
5046 IN int a_attrnamespace;
5047 INOUT struct uio *a_uio;
5049 IN struct ucred *a_cred;
5050 IN struct thread *a_td;
5054 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
5055 struct thread *td = ap->a_td;
5056 struct nameidata nd;
5057 char attrprefix[16];
5058 u_char dirbuf[sizeof(struct dirent)];
5061 struct uio auio, *uio = ap->a_uio;
5062 size_t *sizep = ap->a_size;
5064 vnode_t *xvp = NULL, *vp;
5065 int done, error, eof, pos;
5067 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
5068 ap->a_cred, ap->a_td, VREAD);
5072 error = zfs_create_attrname(ap->a_attrnamespace, "", attrprefix,
5073 sizeof(attrprefix));
5076 plen = strlen(attrprefix);
5083 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
5088 * ENOATTR means that the EA directory does not yet exist,
5089 * i.e. there are no extended attributes there.
5091 if (error == ENOATTR)
5096 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | LOCKSHARED | MPSAFE,
5097 UIO_SYSSPACE, ".", xvp, td);
5100 NDFREE(&nd, NDF_ONLY_PNBUF);
5106 auio.uio_iov = &aiov;
5107 auio.uio_iovcnt = 1;
5108 auio.uio_segflg = UIO_SYSSPACE;
5110 auio.uio_rw = UIO_READ;
5111 auio.uio_offset = 0;
5116 aiov.iov_base = (void *)dirbuf;
5117 aiov.iov_len = sizeof(dirbuf);
5118 auio.uio_resid = sizeof(dirbuf);
5119 error = VOP_READDIR(vp, &auio, ap->a_cred, &eof, NULL, NULL);
5120 done = sizeof(dirbuf) - auio.uio_resid;
5123 for (pos = 0; pos < done;) {
5124 dp = (struct dirent *)(dirbuf + pos);
5125 pos += dp->d_reclen;
5127 * XXX: Temporarily we also accept DT_UNKNOWN, as this
5128 * is what we get when attribute was created on Solaris.
5130 if (dp->d_type != DT_REG && dp->d_type != DT_UNKNOWN)
5132 if (plen == 0 && strncmp(dp->d_name, "freebsd:", 8) == 0)
5134 else if (strncmp(dp->d_name, attrprefix, plen) != 0)
5136 nlen = dp->d_namlen - plen;
5139 else if (uio != NULL) {
5141 * Format of extattr name entry is one byte for
5142 * length and the rest for name.
5144 error = uiomove(&nlen, 1, uio->uio_rw, uio);
5146 error = uiomove(dp->d_name + plen, nlen,
5153 } while (!eof && error == 0);
5162 zfs_freebsd_getacl(ap)
5163 struct vop_getacl_args /* {
5172 vsecattr_t vsecattr;
5174 if (ap->a_type != ACL_TYPE_NFS4)
5177 vsecattr.vsa_mask = VSA_ACE | VSA_ACECNT;
5178 if (error = zfs_getsecattr(ap->a_vp, &vsecattr, 0, ap->a_cred, NULL))
5181 error = acl_from_aces(ap->a_aclp, vsecattr.vsa_aclentp, vsecattr.vsa_aclcnt);
5182 if (vsecattr.vsa_aclentp != NULL)
5183 kmem_free(vsecattr.vsa_aclentp, vsecattr.vsa_aclentsz);
5189 zfs_freebsd_setacl(ap)
5190 struct vop_setacl_args /* {
5199 vsecattr_t vsecattr;
5200 int aclbsize; /* size of acl list in bytes */
5203 if (ap->a_type != ACL_TYPE_NFS4)
5206 if (ap->a_aclp->acl_cnt < 1 || ap->a_aclp->acl_cnt > MAX_ACL_ENTRIES)
5210 * With NFSv4 ACLs, chmod(2) may need to add additional entries,
5211 * splitting every entry into two and appending "canonical six"
5212 * entries at the end. Don't allow for setting an ACL that would
5213 * cause chmod(2) to run out of ACL entries.
5215 if (ap->a_aclp->acl_cnt * 2 + 6 > ACL_MAX_ENTRIES)
5218 error = acl_nfs4_check(ap->a_aclp, ap->a_vp->v_type == VDIR);
5222 vsecattr.vsa_mask = VSA_ACE;
5223 aclbsize = ap->a_aclp->acl_cnt * sizeof(ace_t);
5224 vsecattr.vsa_aclentp = kmem_alloc(aclbsize, KM_SLEEP);
5225 aaclp = vsecattr.vsa_aclentp;
5226 vsecattr.vsa_aclentsz = aclbsize;
5228 aces_from_acl(vsecattr.vsa_aclentp, &vsecattr.vsa_aclcnt, ap->a_aclp);
5229 error = zfs_setsecattr(ap->a_vp, &vsecattr, 0, ap->a_cred, NULL);
5230 kmem_free(aaclp, aclbsize);
5236 zfs_freebsd_aclcheck(ap)
5237 struct vop_aclcheck_args /* {
5246 return (EOPNOTSUPP);
5249 struct vop_vector zfs_vnodeops;
5250 struct vop_vector zfs_fifoops;
5251 struct vop_vector zfs_shareops;
5253 struct vop_vector zfs_vnodeops = {
5254 .vop_default = &default_vnodeops,
5255 .vop_inactive = zfs_freebsd_inactive,
5256 .vop_reclaim = zfs_freebsd_reclaim,
5257 .vop_access = zfs_freebsd_access,
5258 #ifdef FREEBSD_NAMECACHE
5259 .vop_lookup = vfs_cache_lookup,
5260 .vop_cachedlookup = zfs_freebsd_lookup,
5262 .vop_lookup = zfs_freebsd_lookup,
5264 .vop_getattr = zfs_freebsd_getattr,
5265 .vop_setattr = zfs_freebsd_setattr,
5266 .vop_create = zfs_freebsd_create,
5267 .vop_mknod = zfs_freebsd_create,
5268 .vop_mkdir = zfs_freebsd_mkdir,
5269 .vop_readdir = zfs_freebsd_readdir,
5270 .vop_fsync = zfs_freebsd_fsync,
5271 .vop_open = zfs_freebsd_open,
5272 .vop_close = zfs_freebsd_close,
5273 .vop_rmdir = zfs_freebsd_rmdir,
5274 .vop_ioctl = zfs_freebsd_ioctl,
5275 .vop_link = zfs_freebsd_link,
5276 .vop_symlink = zfs_freebsd_symlink,
5277 .vop_readlink = zfs_freebsd_readlink,
5278 .vop_read = zfs_freebsd_read,
5279 .vop_write = zfs_freebsd_write,
5280 .vop_remove = zfs_freebsd_remove,
5281 .vop_rename = zfs_freebsd_rename,
5282 .vop_pathconf = zfs_freebsd_pathconf,
5283 .vop_bmap = VOP_EOPNOTSUPP,
5284 .vop_fid = zfs_freebsd_fid,
5285 .vop_getextattr = zfs_getextattr,
5286 .vop_deleteextattr = zfs_deleteextattr,
5287 .vop_setextattr = zfs_setextattr,
5288 .vop_listextattr = zfs_listextattr,
5289 .vop_getacl = zfs_freebsd_getacl,
5290 .vop_setacl = zfs_freebsd_setacl,
5291 .vop_aclcheck = zfs_freebsd_aclcheck,
5294 struct vop_vector zfs_fifoops = {
5295 .vop_default = &fifo_specops,
5296 .vop_fsync = zfs_freebsd_fsync,
5297 .vop_access = zfs_freebsd_access,
5298 .vop_getattr = zfs_freebsd_getattr,
5299 .vop_inactive = zfs_freebsd_inactive,
5300 .vop_read = VOP_PANIC,
5301 .vop_reclaim = zfs_freebsd_reclaim,
5302 .vop_setattr = zfs_freebsd_setattr,
5303 .vop_write = VOP_PANIC,
5304 .vop_pathconf = zfs_freebsd_fifo_pathconf,
5305 .vop_fid = zfs_freebsd_fid,
5306 .vop_getacl = zfs_freebsd_getacl,
5307 .vop_setacl = zfs_freebsd_setacl,
5308 .vop_aclcheck = zfs_freebsd_aclcheck,
5312 * special share hidden files vnode operations template
5314 struct vop_vector zfs_shareops = {
5315 .vop_default = &default_vnodeops,
5316 .vop_access = zfs_freebsd_access,
5317 .vop_inactive = zfs_freebsd_inactive,
5318 .vop_reclaim = zfs_freebsd_reclaim,
5319 .vop_fid = zfs_freebsd_fid,
5320 .vop_pathconf = zfs_freebsd_pathconf,